System and methods for evaluating drug libraries in high throughput

Abstract

Disclosed are compositions, devices, systems, and methods for used for synthesizing peptides, proteins, antibodies, drug molecules or candidates, and biologies (biologic drug molecules), to provide on-chip protein or drug or biologies libraries, for drug discovery screening or diagnostics purposes.

Description

[0001] Atty. Dkt. No. 155529.00033

[0002] SYSTEM AND METHODS FOR EVALUATING DRUG LIBRARIES IN HIGH THROUGHPUT CROSS-REFERENCE TO RELATED APPLICATIONS This patent application claims the benefit of priority to United States Provisional Patent Application No. 63 / 824,525, filed June 16, 2025, and Provisional Patent Application No.

[0003] 63 / 744,256, filed January 11, 2025, each of which is herein incorporated by reference in its entirety.

[0004] SEQUENCE LISTING

[0005] The contents of the electronic sequence listing (15552900033_SL.xml; Size: 6,333 bytes; and Date of Creation: January 12, 2026) are herein incorporated by reference in their entirety7.

[0006] BACKGROUND

[0007] Accurate detection of biomolecular interactions is essential in many areas, from detecting the presence of biomarkers in the clinic, development of therapeutic drugs and biologies in biopharma, to understanding various biological processes in basic research. Traditional endpoint approaches can suffer from false-negative results for biomolecular interactions with fast kinetics, or false positives due to non-specific interactions. By contrast, real-time detection techniques like surface plasmon resonance (SPR) monitor interactions as they form and disassemble, reducing the risk of false-negative results. Real-time sensor-integrated proteome on chip (SPOC®) SPR-based detection can improve real-time biomarker screening and kinetic evaluation.

[0008] Drug discovery7continues to face a staggering 90% failure rate, with many setbacks occurring during late-stage clinical trials. To address this challenge, there is an increasing focus on developing and evaluating new technologies to enhance the "design" and "test" phases of antibody-based drugs (e.g., monoclonal antibodies, bispecifics, CAR-T therapies, ADCs) and biologies (biologic drugs or biologic drug molecules) during early preclinical development, with the goal of identifying lead molecules with a higher likelihood of clinical success. Artificial intelligence (Al) is becoming an indispensable tool in this domain, both for improving molecules identified through traditional approaches and for the de novo design of novel therapeutics. However, critical bottlenecks persist in the "build" and "test" phases of AI-designed antibodies and protein or peptide binders, impeding early preclinical evaluation. While Al models can rapidly generate thousands to millions of putative drug designs.Atty. Dkt. No. 155529.00033

[0009] technological and cost limitations mean that only a few dozen candidates are ty pically produced and tested.

[0010] Sensor-integrated Proteome On Chip (SPOC®) platform enables the production and capture-purification of 1,000 - 2,400 folded proteins directly onto a glass slide or any biosensor or surface plasmon resonance (SPR) biosensor chip for measuring kinetic binding rates with picomolar affinity resolution. SPOC technology is used to express single-chain antibodies (sc-antibodies), specifically scFv, VHH, dual chain Fab constructs and full-length antibodies or multi-specifics, among others. These constructs are capture-purified at high levels on SPR biosensors and retain functionality' by the binding specificity7to their respective target antigens (introduced as analyte in solution). SPOC outputs comprehensive kinetic data including quantitative binding (Rmax), on-rate (ka). off-rate (kd). affinity (KD), and half-life (t 1 / 2). for each of thousands of on-chip antibody fragments or proteins or peptides or biologies.

[0011] Cell-free protein expression leverages carefully prepared lysates from various cultured bacterial, prokary otic or mammalian, insect, plant and other eukaryotic cells to achieve in vitro transcription and translation (IVTT) of proteins. Like any metabolic process. IVTT consumes fuel and produces reaction by-products, both of which can drastically impact the output and efficiency of cell-free protein expression. A proper concentration and balance of inputs, e g. mRNA and protein precursor molecules like nucleotide triphosphates and amino acids respectively, enzymatic co-factors like magnesium and potassium, and energy sources like adenosine triphosphate (ATP) and guanosine triphosphate (GTP), among many other factors, are critical to maximize the functional protein yields achievable across various cell-free systems.

[0012] In certain embodiments, the SPOC (Sensor-integrated Proteome-On-Chip) platform enables high-throughput expression, folding, modification, and quantitative biophysical characterization of proteins, peptides, antibodies, and other biologically active constructs directly on a biosensor surface. The platform combines cell-free protein synthesis with realtime sensing technologies (e.g., surface plasmon resonance, multi -parameter optical sensing, or mass-spectrometric interrogation) to establish a single integrated system capable of performing numerous drug-discovery assay types. These include affinity and kinetics assays, structure-function mapping assays, epitope and variant scanning assays, multi-target profiling assays, enzyme activity7assays, and ALdirected design-test-leam optimization cycles.Atty. Dkt. No. 155529.00033

[0013] SUMMARY

[0014] One aspect of the disclosure is a system for integrated on-chip drug discovery’ of peptide- or protein-based drug molecules or their conjugates, the system comprising:

[0015] a) a nanowell array comprising: a nanowell slide comprising a plurality of wells; and a biosensor surface comprising an array of discrete locations, wherein a plurality7of drug molecules is captured to the biosensor surface to generate a plurality7of on-chip drug molecules, wherein each discrete location comprises one drug molecule of the plurality of drug molecules captured to the biosensor surface, wherein the plurality of drug molecules comprises one or more peptide, or protein, wherein at least a portion of the plurality of drug molecules is configured to be generated on the biosensor surface from nucleic acid sequences using in vitro cell-free or cell-based expression system in the plurality7of wells; and wherein the plurality7of drug molecules is configured to bind covalently to the biosensor surface and form the plurality of on-chip drug molecules;

[0016] b) a fluidic module configured to deliver one or more analyte or target molecule to the plurality7of on-chip drug molecules under controlled assay conditions, wherein the plurality7of on-chip drug molecules is configured to be regenerated after assaying with the one or more analyte or target molecule, for multiple successive assay and regeneration cycles; and

[0017] c) a detection or sensing module configured to measure binding interactions between the plurality7of on-chip drug molecules and the one or more analyte or target molecule, wherein the system is configured to perform synthesis or presentation of the drug molecules and measurement of binding interactions on the same solid support surface without transferring the drug molecules off the biosensor surface.

[0018] A second aspect is a method for integrated on-chip drug discovery of peptide- or protein-based drug molecules or their conjugates, the method comprising: (a) synthesizing a plurality of drug molecules and attaching the plurality of drug molecules to a solid support surface at discrete locations on the solid support surface, wherein the plurality of drug molecules are synthesized from nucleic acid sequences in cell-free systems or cell-based systems in sealed wells, wherein the plurality7of drug molecules bind covalently to the solid support surface and form a plurality of on-chip drug molecules; (b) contacting the plurality of on-chip drug molecules on the solid support wi th one or more target molecules under controlled assay conditions; (c) measuring binding interactions between the drug molecules and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociationAtty. Dkt. No. 155529.00033

[0019] constant (KD), or mass spectrometer data, or combinations thereof; and (d) analyzing the binding data to perform one or more drug discovery operations selected from affinity ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping, wherein the generating of the drug molecules and the measuring of the binding interactions are performed on the same solid support surface in an integrated workflow.

[0020] Another aspect is a method of integrated on-chip drug discovery of peptide- or protein-based drug molecules or their conjugates, the method comprising: a) contacting a plurality of on-chip peptide- or protein-based drug molecules attached covalently to a solid support with one or more target molecules, wherein the on-chip drug molecules are soluble and capable of binding their native ligand; b) measuring binding interactions between the drug molecules and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or mass spec data, or combinations thereof; and c) analyzing the binding data to perform one or more drug discover)’ operations selected from affinity ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping.

[0021] Another aspect is a method of integrated on-chip drug discovery of peptide- or protein-based drug molecules or their conjugates, the method comprising: a) contacting a plurality of on-chip peptide- or protein-based drug molecules attached covalently to a solid support with one or more target molecules; b) measuring binding interactions between the drug molecules and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or mass spec data, or combinations thereof; c) analyzing the binding data to perform one or more drug discovery operations selected from affinity ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping; d) stripping the one or more target molecules from the plurality of drug molecules while leaving the drug molecules covalently bound to the solid support surface; and e) repeating steps a) through c) with a second target molecule.Atty. Dkt. No. 155529.00033

[0022] BRIEF DESCRIPTION OF THE FIGURES

[0023] The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0024] It should be understood that the drawings described below are for illustration purposes only and are not intended to limit the scope of the claims in any way.

[0025] FIGS. 1A-1E depicts SPOC drug discovery workflow from candidate binder sequence (e.g. peptide / protein / Ab sequences) (FIG. 1A). DNA synthesis (FIG. IB), SPOC SPR chip generation via cell-free synthesis (FIG. 1C), and SPR screening of all sequences (FIG. ID) simultaneously. SPR screening provides full SPR kinetics data for the entire on-chip drug library (FIG. IE). Covalent capture mechanism enables regeneration and re-use of the same sensor chip for additional follow-on assays which inform therapeutic candidate selection or deselection.

[0026] FIGS. 2A-2B. FIG. 2A illustrates one example method for preparing a protein array on biosensor surfaces 200 via cell-free protein expression and in situ capture-purification onto biosensor surfaces 200. Nanowell slides 604 containing linear or plasmid DNA 612a-c harboring tagged proteins of interest act as nano-liter volume reaction vessels or micro reaction chambers for transcription and translation by cell-free protein expression lysate 705. The nanowells 608a-c are sealed after being filled with the cell-free protein expression lysate 705 by a biosensor surface 200 or glass capture slide, creating closed or sealed nano-liter or microliter volume reaction arrays, which after a 2-hour incubation yields an array or on-chip library of purified expressed proteins 717 that have been captured on to the biosensor surface 200 at discrete spots located directly above each nanowell 608a-c on the surface in an array format, overcoming the need for individual protein purification in cellular systems or automated robotic spotting. The incubation time for protein expression can be as low as 15 minutes and as high as 24 or 48 hours or higher. FIG. 2B depicts nanowells or nano-liter volume wells. SEM image of 375 pm (period). FIG. 2C depicts nanowells filled with cell-free protein expression lysate. FIG. 2D depicts expressed protein captured on a second surface or biosensor surface via capture tag.

[0027] FIG. 3 depicts an exemplary SPOC SPR chip with a plurality of biomolecules (shown with VHH as an example) bound covalently on the surface (shown with Halotag as an example). A target biomolecule (shown as a protein as an example), or off-target, is flowed across the surface wherein kinetic features of the interaction, or lack of interaction, areAtty. Dkt. No. 155529.00033

[0028] measured and used to subsequently rank the binders to down-select candidates for further screening and development.

[0029] FIG. 4 depicts an exemplary SPOC SPR chip with a plurality of biomolecules (show n with VHH as an example) bound covalently on the surface (shown with Halotag as an example). A target biomolecule (show n as a protein as an example), or off-target, is flowed across the surface followed by flowing another biomolecule binder (shown as VHH as an example) across the sensor surface. If the biomolecule flowed across the surface shares an epitope with the covalently bound binder on the surface, then no interaction is expected. If the biomolecule flowed across the surface targets a unique epitope, then interaction is expected to occur. The entire covalently bound library can be placed into bins in this manner by repeatedly flowing additional biomolecules and assessing for competitive interactions for library-scale epitope binning analysis.

[0030] FIGS. 5A-5C depict an exemplary library-scale epitope binning analysis dataset. Three antibodies each targeting a unique epitope on HER2 protein were individually and sequentially flowed across the surface after allowing HER2 protein to interact with the covalently bound biomolecule library on the SPOC chip (FIG. 5A). Data clearly shows the anti-HER2 VHH did not overlap epitopes with Trastuzumab or Pertuzumab, as expected, while Trastuzumab only shows competition with itself, as expected (FIG. 5B). Rituximab targets CD20 and was included as a negative control. Data is collected on 1,000+ biologies candidates on-chip simultaneously for library-scale epitope binning (FIG. 5C).

[0031] FIG. 6. Exemplary SPOC SPR chip with a plurality of biomolecules (shown with VHH as an example) bound covalently on the surface (shown with Halotag as an example). Target biomolecules (shown as nucleic acids and proteins, as examples), are flowed across the surface one by one, sequentially, wherein kinetic features of the interaction, or lack of interaction, are measured and used to subsequently down-select candidates for further screening and development. This assay is a de-selection example step wherein binders with high binding to one or more polyreactive reagents are removed from consideration for downstream therapeutic candidate development.

[0032] FIG. 7. Two antibodies which both target Halotag protein were screened on a SPOC chip via end-point fluorescent and real-time SPR methods. The Mouse anti-HaloTag antibody on the left shows minimal fluorescence, while the Rabbit anti-HaloTag antibody on the right shows extensive fluorescent signal, while both antibodies had detectable binding events across the array when measured via SPR, suggesting that the antibody on the left had missed interactions with proteins on the surface in the fluorescent assay. Analysis of the kinetic profilesAtty. Dkt. No. 155529.00033

[0033] of each antibody demonstrates that the fast off-rate of the antibody shown on the left results in washing off during fluorescent assay wash steps (required of all endpoint style assays), and therefore may preclude detection of the interaction. Current fluorescence and other end-point methods miss fast-off rate binders due to wash steps. Off-target interactions often result in toxicity, and may lead to failures in Phase I. SPOC SPR real-time biosensing detects fast off-rate binders, which if unaddressed can result in toxicity during Phase I safety trials. This demonstrates the utility of SPOC SPR chips for detecting fast off-rate (transient) binding events, which may be critical for early detection of potential off-target binding resulting in later toxicity7and failure in clinical trials.

[0034] FIG. 8 depicts SPOC areas of applications in the study of proteins and proteoforms at scale, disease science, drug discovery and development, plasma proteomics, and diagnostics.

[0035] FIG. 9 depicts exemplary biosensors based on the type of biosensor, such as receptor based, electrochemical, optical, piezoelectric, and nanobiosensors.

[0036] FIGS. 10A-10C illustrate the plurality7of capture surfaces able to be employed for ligand capture and screening by various end-point and real-time assays as outlined in this disclosure. Two exemplary capture surface modalities are demonstrated for screening by three readouts via fluorescence, surface plasmon resonance, and matrix-assisted laser desorption / ionization (MALDI) mass spectroscopy. FIG. 10A: top: glass capture surface; bottom: interaction readout. FIG. 10B: top: gold biosensor surface, SPR real-time assay; bottom: interaction readout. FIG. 10C: top: gold biosensor surface, MALDI imaging end point assay; bottom: interaction readout of MBP probe binding signal and MBP probe peak.

[0037] FIGS. 11A-11C depict a biosensor chip as viewed by SPR instrument showing different capture densities 1,000 and 2,400 (FIG. 11 A) and simultaneous collection of qualitative, quantitative, and kinetic data across the entire biosensor chip as sensograms each representing data from a unique spot on the biosensor (FIG. 1 IB), with expanded view of an anti-capture tag antibody titration data with mouse anti-capture tag (FIG. 11C). These show the expansion capability7to larger numbers of targets with the SPOC array, with little to no cross-talk as indicated by antigen-specific and capture-tag antibodies probing the same sensor locations.

[0038] FIG. 12 is a graph illustrating an example sensogram or kinetic trace from which one or more kinetic parameters of protein or drug binding can be derived after proper controlsurface normalization and blank-injection referencing. The parameters can include the association and dissociation rates, the max signal observed, binding affinity,, residence time.

[0039] FIGS. 13A-13B depicts an array of single chain antibody fragments produced on SPOC chip (FIG. 13A; target binding site on an antibody (paratope)) and an example of a SPOC chipAtty. Dkt. No. 155529.00033

[0040] with full length folded proteins and antibody fragments (FIG. 13B). scFv means single-chain variable fragment, VH means variable heavy domain, VL means variable light domain, and CDR1-3 are complementarity determining regions. The antibody array chip can be used to screen with therapeutic target and down-select antibodies with optimal kinetic binding parameters and sequence diversity for therapeutics development. The antibody array chip can be used to identify and characterize target binding site on an antibody (paratope) with single amino acid resolution and kinetic data. To identify the specific amino acids in the antibody CDR regions that are critical to binding the target or analyte, the amino acids in the CDR regions and the neighboring regions are mutationally scanned, by replacing one or more amino acids, sequentially or tiled, with alanine and / or other amino acids. The mutationally scanned antibody chip is then screened with target or analyte, and change in binding characteristics is measured. Loss or gain of binding (or binding affinities) caused by the specific mutations indicate the amino acids critical to binding (paratope) the target. To produce antibody sequences with high or optimal binding parameters or therapeutic development properties, mutations are induced in the antibody CDR and neighboring regions to create large mutational antibody libraries on SPOC chip. Mutations can be induced using in-silico modeling, ML or Al drug design, or randomly, or using other approaches. By screening with target or analyte, the antibodies sequences with high or optimal target binding parameters or therapeutic development properties can be down selected.

[0041] FIGS. 14A-14B depicts on-chip synthesized array of antibodies or antibody fragments for plasma proteomics. Each spot is an immunoglobulin-like protein (Ig-like) (antibody fragment or full-length antibody) representing an scFv, nanobody - VHH, VH, or Fab (FIG.

[0042] 14A) or a full length antibody, mammalian antibody, or camelid antibody (FIG. 14B) that is designed to bind to a specific plasma protein or biomarker or antigen or metabolite or nucleic acid or biomolecule or toxin or environmental analyte or ion or organic molecule.

[0043] FIG. 15 is a map of the commercial T7 Cell-free Expression Vector. The Custom SPOC pT7CFEl_nHalo_cHis vector was derived from this commercial vector for cell-free protein expression.

[0044] FIGS. 16A-16D depicts an exemplary process for identifying critical epitope residues on a structure. The exemplary process may comprise comprehensive amin acid scanning across protein(s) (FIG. 16A), expression in nano wells and capture on biosensors (FIG. 16B), comparison of binding data across mutated variants (FIG. 16C), and identification of critical epitope residues on structures (FIG. 16D). The process may be used for epitope identification, vaccine candidate identification / protein engineering, and / or antibody therapeutic development.Atty. Dkt. No. 155529.00033

[0045] FIG. 17 is a schematic of current antibody drug development approaches and the proposed integration of SPOC into these workflows. Antibody sequence libraries have traditionally been harvested from 1) B-cells collected from naive or immunized animals or convalescent patient serum, 2) derived from synthetic cDNA constructs, and 3) increasingly from generative artificial intelligence (Al) applied to in silico design of biologies. These inputs feed into a plurality of screening assays. Isolated B-cells are cultured, and secreted antibodies are assayed for clones that bind target antigen, which are subsequently sequenced. For inputs where sequences are already known, as after B-cell sequencing or for synthetic libraries or Al, various display technologies (phage, yeast, and ribosomes) can be directly employed whereby these cell-lines or ribosomes are applied to synthesize millions of sc-antibody variants for downstream panning against target antigens, and sequencing of the binders. After initial panning assays that isolate and identify the binders, deeper characterization is performed to assess the functional activity of lead candidates and interrogate their target binding mechanisms and kinetics with the use of label-free techniques such as surface plasmon resonance (SPR). These workflows involve additional paratope characterization steps and affinity maturation cycles, with the objective of down-selecting final lead sequences that are further engineered and optimized for efficacy and safety, prior to initiating humanization, Fc engineering and in vivo studies. SPOC platform can integrate into current workflow s (low er part of the figure), wherein full length or variable domain sequences derived from libraries can be directly leveraged for cell-free production and capture of antibody fragment libraries on SPR biosensor chips. SPOC facilitates measuring binding kinetics of all binders for down-selection by producing rich data, enabling deep characterization of the paratopes of lead antibodies and antibody affinity maturation cycles towards selecting optimally engineered lead candidates.

[0046] FIG. 18 is a schematic of SPOC assay integration into iterative Al-driven discovery w orkflows. Using inputs from trained Al models, sequences for numerous antibody sequences or protein binders (or other target-binding scaffolds and protein designs) can be cloned into cell-free expression vectors to rapidly build sc-antibody or protein libraries on SPOC SPR chips that can be screened with the desired target(s) to yield a wealth of detailed kinetic binding data. These rich kinetic datasets can then feed back into Al models providing additional training data to learn from, enabling the further refinement of Al-designed binders.

[0047] FIGS. 19A-19D depicts full length antibodies (FIG. 19A), single chain variable fragment (scFv) of antibody (FIG. 19B), dual chain Fab antibody fragment (FIG. 19C), and Full length antibody, various single chain fragments. scFv, and Fab proteins shown with post-translational modifications (FIG. 19D).Atty. Dkt. No. 155529.00033

[0048] FIG. 20 depicts different antibody or protein-based drug designs, each of which can be produced in exemplary embodiments in nanowells and captured on chip.

[0049] FIGS. 21A-21B depicts exemplary membrane proteins with example disulfide PTMs, each of which can be produced in exemplary embodiments in nanowells and captured on chip.

[0050] FIG. 22 depicts SDS-PAGE analysis of C-Tertninal HaloTagged scFvs and VHH expressed using E. coli lysate in sealed nanowells. All expressed constructs were first covalently labeled with fluorescent Halo ligand then analyzed in reducing (R) and non-reducing (NR) conditions. A subset of constructs were expressed before (-Opt) and after (+Opt) optimization.

[0051] FIG. 23 depicts confirmation of capture of scFv and VHH proteins with C-terminal HaloTag on SPOC biosensor. Capture of single chain antibodies on the biosensor following expression via E. coli IVTT lysate was validated via anti-HaloTag antibody via SPR.

[0052] FIGS. 24A-24D. Analysis of scFv and VHH proteins on SPOC biosensor binding to antigen targets in comparison to non-targets. Antigen targets of the scFv and VHH proteins were analyzed for specific binding against all scFv and VHH simultaneously using a SPOC SPR biosensor. Binding traces to each cognate single chain antibody are shown in FIG. 24A, while maximum RU signal measured against all constructs at the end of the association phase (Rviax) is displayed in FIG. 24B. Binding traces of sandwdch antibodies against each antigen, measured after flowing the antigen across the sensor, were collected and are shown in FIG. 24C, while the R\iax for each is shown in FIG. 24D. IgG / IgM-Depleted Serum was used as the analyte for HAS. X indicates selectivity data was not collected as part of this data set.

[0053] FIGS. 25A-25B. Kinetic analysis of TNFa and HER2 VHH constructs via SPR. Affinity constants for interactions between anti-TNFa VHH and TNFa w ere calculated from a titration of TNFa binding to the SPOC SPR sensor (FIG. 25 A). Regenerative conditions were used between TNFa injections due to high affinity of the interaction. However, lack of observed dissociation in 30-minute window' of the experiment (data not shown) indicates that the affinity measured is an approximation of the real value, and disassociation measurements over significantly longer periods will be needed for accurate pM affinity calculation. The affinity constant for the interaction between anti-HER2 VHH with recombinant HER2 (FIG. 25B) w as calculated from a titration of HER2 extracellular domain (ECD) binding using non-regenerative conditions. Kinetic measurements were averaged over the duplicate spots (1 and 2) for both VHH constructs.

[0054] FIGS. 26A-26B. HER2 VHH (2Rsl5d) wildtype sequence and resulting substitutions used in the mutation study. The sequence of HER2 VHH (SEQ ID NO: 2) and identified CDRs of HER2 VHH (2Rsl5d) (SEQ ID NOs: 3-5) (FIG. 26A). CDRs are highlighted and amino acid positionsAtty. Dkt. No. 155529.00033

[0055] indicated for each. CDR amino acid substitutions to alanine, aspartate, lysine, and serine were designed into the constructs, resulting in 92 variants as shown in FIG. 26B.

[0056] FIG. 27. Select sensorgrams showing mutants with higher affinity compared to WT sequence. Left: Affinity curves (red) fitted to entire curve; Right: Curves fitted only for the dissociation (red)

[0057] FIGS. 28A-28B depicts sensorgrams showing HER2 ECD titration kinetics for a single replicate each of the 92 HER2 VHH mutants. Titration was performed using regenerative conditions with increasing concentrations of HER2 ECD (1.4 nM, 4.1 nM, 12 nM, 37 nM, 110 nM), with all injections overlayed onto a single sensorgram. The red lines indicate the kinetic model curve fits to underlying data plots.

[0058] FIG. 29 depicts a visualization of residues that resulted in loss of binding or higher affinity upon substitution, in comparison to wildtype. CDRs 1, 2, and 3 are denoted in light yellow, light orange, and light green, respectively. Residues which produced higher affinity upon substitution are summarized visually in blue for each amino acid substitute (alanine, Ala; lysine, Lys; serine, Ser; aspartate, Asp) and specific substitutions are noted for each. Residues which resulted in loss of binding (red) or lower affinity 3-times or greater compared to wildtype (pink) are visualized in the lower left comer, with specific substitutions that resulted in loss of binding noted for each.

[0059] FIG. 30 depicts results showing that varying input DNA concentration results in similar affinities for expressed protein. DNA encoding HER2 VHH wildtype was titrated during the nanowell printing process and the resulting protein expressed and captured onto the SPOC sensor was evaluated for binding with HER2 ECD.

[0060] FIGS. 31A-31C depicts bar graphs showing kinetic data obtained following a binding study of HER2 ECD against HER2 VHH mutants using a titration of HER2 ECD, organized by CDR and kinetic parameter. Affinity constant KD for each amino acid substitution is shown in FIG. 31 A, on-rate ka is shown in FIG. 3 IB, and off-rate kd is shown in FIG. 31C. Missing bars indicate no binding was observed as defined in the methods.

[0061] FIGS. 32A-32D depicts demonstration of epitope binning using SPOC. (FIG. 32A) Schematic of HER2 depicting the four extracellular domains (ECD) and the non-overlapping domains targeted by the therapeutic antibodies and anti-HER2 VHH used in this study. The anti-HER2 VHH (orange) binds ECD 1, Pertuzumab (green) binds ECD 2, and Trastuzumab (pink) binds ECD 4. (FIG. 32B) Graphical depiction of the analyte injections involved in this binning experiment. Sensor bound HaloTagged (blue circle) anti-HER2 VHH (orange) andAtty. Dkt. No. 155529.00033

[0062] HaloTagged Trastuzumab scFv (pink) are depicted bound to the SPOC sensor surface. After HER2 ECD (light blue) injection and pre-binding, the sandwich antibody (Trastuzumab in this example) is injected and will bind any HER2 ECD bound to ligand with non-overlapping epitopes (i.e., anti-HER2 VHH). After antibody injection, the sensor is regenerated and HER2 ECD is re-injected and the cycle repeats for additional antibodies to be screened. (FIG. 32C) Sensorgram traces from sensor bound HaloTagged anti-HER2 VHH with injected Trastuzumab and Pertuzumab (green highlighted area) after HER2 ECD pre-loading (red space). (FIG. 32D) Sensorgram traces from sensor bound HaloTagged anti-HER2 Trastuzumab scFv binding HER2 ECD in the first injection, and then following Trastuzuman, Pertuzumab, or Rituximab injections.

[0063] FIGS. 33A-C depicts demonstration of epitope binning using IVTT lysate as an antibody analyte source as opposed to purified antibody. (FIG. 33 A) Graphical depiction of the analyte injections involved in this binning experiment. Sensor bound HaloTagged (blue circle) anti-HER2 VHH (orange) and HaloTagged Trastuzumab scFv (pink) are depicted bound to the SPOC sensor surface. After HER2 ECD (light blue) injection and pre-binding, diluted IVTT lysate containing free-floating HaloTagged anti-HER2 VHH or HaloTagged Trastuzumab scFv is injected to interrogate capability for productive sandwich formation, an indication of nonoverlapping HER2 ECD binding sites with the corresponding ligand on the sensor surface. The sensor was regenerated and cycle repeated for subsequent HER2 ECD and IVTT lysate injections. (FIG. 33B) Injection of IVTT lysate expressing HaloTagged anti-HER2 VHH was performed after HER2 ECD preloading on a sensor containing both HaloTagged anti-HER2 VHH (orange trace) and HaloTagged Trastuzumab scFv (purple trace). Productive sandwich (green highlighted area) was only observed between IVTT anti-HER2 VHH and sensor bound HaloTagged Trastuzumab scFv. (FIG. 33C) Same experimental setup as in FIG. 33B but with IVTT HaloTagged Trastuzumab scFv sample used as analyte, only generating productive sandwich with sensor bound HaloTagged anti-HER2 VHH (green highlighted area) as expected.

[0064] FIG. 34 depicts exemplary SPOC SPR chip with a plurality of biomolecules (shown with VHH as an example) bound covalently on the surface (shown with Halotag as an example). The VHH molecules on the surface have been mutationally scanned, substituting one or more different amino acids at each CDR residue to create a mutationally scanned library7on-chip in an attempt to generate new binders with a desired kinetic or affinity, profile. A target biomolecule (shown as a protein as an example), or off-target, is flowed across the surface wherein kinetic features of the interaction, or lack of interaction, are measured and used toAtty. Dkt. No. 155529.00033

[0065] subsequently rank the binders to down-select candidates for further screening and development, or identify the final lead candidate(s).

[0066] FIGS. 35A-35D. Data is shown with an anti-HER2 VHH mutationally scanned with 4 amino acid substitutions at each CDR residue, one amino acid to represent each class of amino acids (Alanine: non-polar hydrophobic; Aspartate: acidic; Lysine: basic; Serine: neutral, polar) (FIG. 35 A). This resulted in ~96 mutational anti-HER2 VHH variants which were synthesized on SPOC SPR chips and screened vs HER2 and kinetic features measured via SPR (FIG. 35B). Clones were then ranked based on different kinetic features (KD, ka, kd, tl / 2, etc.) to identify those with improved features or worsened features, such as KD (FIG. 35C). KD values were mapped onto the known VHH structure to identify' CDR regions which resulted in gain or loss of affinity (FIG. 35D).

[0067] FIG. 36 depicts synthesis on chip along with its deep mutationally scanned variants, such as single amino acid alanine substitutions, and lead drug molecules are screened as analyte, for high resolution epitope mapping with single amino acid resolution.

[0068] FIG. 37 SPOC mutationally scanned biomolecule library chips can be used to generate library-scale kinetics data for use in antibody patent applications, as now required due to recent court rulings. SPOC data can be used in FDA filings and patent filings to protect biologies assets with sequences containing desired features of the final candidate.

[0069] FIG. 38 is a schematical overview of how nanodiscs are formed by two components: a lipids, such as l,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), and structural peptides (e.g., ApoAl ). Here, an expression plasmid for the structural peptide and the appropriate lipids are mixed with cell-free protein lysate which expresses the ApoAl. ApoAl wraps around the lipids forming a raft structure that yields a nanodisc of defined shape and size which can be used for stabilizing membrane proteins for downstream screening.

[0070] FIG. 39 is an illustration incorporating membrane nanodisc technology with SPOC platform, to produce and capture full-length functional membrane proteins with transmembrane domains on SPR biosensor chips. Membrane Nanodiscs enable full-length membrane protein expression and assembly in cell based and cell-free protein expression systems. Co-express membrane proteins and nanodiscs in nanowells and capture on SPR biosensor chip. Nanodiscs are co-expressed in SPOC nanowells by dispensing plasmid-DNA for membrane-scaffolding-proteins (such as but not limited to) human apolipoprotein Al along with plasmid-DNA of target membrane proteins, and adding lipid vesicles (neutral or partially anionic or cationic) or lipid reagents and other reagents to cell-free lysate system. Co-expressed and assembled membrane protein-nanodisc complexes, in respective nanowells, are captured as array-spotsAtty. Dkt. No. 155529.00033

[0071] onto SPOC chips via HaloTag fusion protein, that is appended either to apolipoprotein or c-terminus of membrane proteins.

[0072] FIGS. 40A-40C. Gel analysis of Humira Fab and scFv constructs. FIG. 40A shows that initial testing indicated that the Fab fragments were soluble, but scFv was aggregating significantly. Further testing oftheFab constructs indicated that a 1:1 molar ratio of heavy: light chain was likely most ideal (FIG. 40B). An attempt to increase the soluble fraction of scFv via reduction of DNA input indicated that the soluble fraction remained consistent while aggregated fraction reduced with a DNA input 100-less than standard input (10 nM). A small DOE was performed to find the best expression conditions for improving the scFv soluble fraction. It was determined that expression at 26 C was key to allowing Humira scFv to properly fold, without significant insoluble protein signal (FIG. 40C).

[0073] FIG. 41 depicts an exemplary aggregation study design: His-tagged HC vs. His-Tagged LC. HaloTag is indicated by a star and his-tag is indicated by a circle.

[0074] FIG. 42 depicts Humira expression conditions vs SPR functionality'. Binding was not assessed, and thus no binding was detected in the lop-left panel. However, binding was detected in the other three panels.

[0075] FIG. 43 depicts a functional TNFa binding observations overview, with a focus on Adalimumab. The table show s the ty pe and concentration of constructs used vs the temperature. Graphs w ere then developed based on the time vs the response.

[0076] FIG. 44A-44B. Fig. 44 A shows the gel electrophoresis under reducing and non-reducing conditions of Trastuzumab Fab, monoclonal antibody and scFv constructs when expressed under similar conditions. Fig. 44B shows SPR sensorgram data for Trastuzumab Fab, mAb, and scFv formats binding to various ligands when expressed under similar conditions. Similar performance across constructs is demonstrated when given the same expression conditions.

[0077] FIGS. 45A-45G depicts the effect of acidic buffer regeneration treatment on detection of Antibody #1 and #2 binding. A glass microscope slide with arrayed HaloTag protein was prepared and probed with Antibody #1 and subsequent Cy3-labeled goat anti mouse secondary antibody for detection before (FIG. 45 A) or after (FIG. 45B) the arrays were exposed to a 5.0 minute acid wash with 10 mM Glycine-HCl (pH = 2.4). Sensor gram of binding signal obtained from 355 spots across a HaloTag ligand spotted biosensor surface over the course of Antibody #1 injection before (FIG. 45C) or after (FIG. 45D) the biosensor has been regenerated with the 10 mM Glycine-HCl (pH = 2.4) acidic regeneration buffer. Sensorgram of binding signal obtained from 355 spots across the biosensor surface over the course of Antibody #2 injection before (FIG. 45E) or after (FIG. 45F) the biosensor had been regenerated. Each kinetic trace inAtty. Dkt. No. 155529.00033

[0078] all sensorgrams represents binding signal observed from a single ROI selected for analysis on the SPOC array.

[0079] FIGS. 46A-46C shows that covalent capture of HaloTag ligands enables SPOC biosensor regeneration without loss of protein ligand. These experiments w ere performed on a biosensor which was already exposed to the acid regeneration buffer, so no increase in Antibody #1 binding signal was expected. (FIG. 46 A) Sensorgram of ten Antibody #1 serial injections (blue traces) were performed with 60 seconds of 10 mM Glycine-HCl (pH = 2.4) acid regeneration (red bars) between each antibody injection. Complete return of Response (RU) signal back to baseline after acid regeneration indicates that all bound antibody was successfully stripped prior to subsequent replicate injections. (FIG. 46B) Sensorgrams of Antibody #1 binding signal before (blue) and after (red) 35 cycles of acid regeneration. (FIG.

[0080] 46C) Scatter plot comparing Antibody #1 binding signals measured from panel B between initial (blue) and final (red) antibody injections, following 35 acid regeneration cycles. Y-axis plots Rmax values of the first Anti -463 body #1 injection, and the X-axis plots Rmax values of the final Antibody #1 injection.

[0081] FIG. 47. Full scan of the IVTT HaloTag fusion protein spotted glass slide probed with anti-HaloTag Antibody #1 (panel on the left) and Antibody #2 (panel on the right). Subarrays #1 and #2 are displayed in the area contained by the dashed white box. Each of the four subarrays included in this analysis contain two replicate sets (A and B) printed to enable capture of IVTT expressed HaloTag protein ligands within sparsely and a densely spotted region. The same glass slide was re-probed directly with Antibody #2 primary after initially probing with Antibody #1, without exposing the capture surface to any stripping or regeneration buffer solution. The array in the center of the slides w ere excluded from consideration, though the same disparity between the antibody binding patterns emerges.

[0082] FIGS. 48A-48B. Individual sensorgrams of binding signal at each HaloTag protein ligand from the entire 355 ROI array (FIG. 48A). The kinetic data at each ROI is shown, with Antibody # 1 binding signal colored blue and Antibody #2 binding signal colored green. The 1: 1 binding kinetic model used for obtaining off-rate (kd) parameters is overlaid on the traces and indicated by the red line. A representative sub-set of the data is blown up larger beneath the 355 sensorgrams and is indicated by the dashed outline (FIG. 48B).

[0083] FIGS. 49A-49B depicts thresholds used for obtaining Rmax values for ligand binding comparative analysis. Sensorgrams of binding signal obtained from 355 spots across the biosensor surface during Antibody #1 injection pre- (FIG. 49A) or post- (FIG. 49B) acid regeneration with 10 mM Glycine-HCl (pH=2.4). Average signal measured within the boundsAtty. Dkt. No. 155529.00033

[0084] of the thresholds (pink vertical lines) were collected for each ligand, during both antibody injections, and utilized to generate the scater plot in FIG. 50A. Similar thresholds were set for Antibody #2 and used for the scater plot shown in FIG. 50B.

[0085] FIGS. 50A-50B depicts effect of acidic regeneration on anti-HaloTag binding signal. Scater plot comparing Rmax values measured from 355 spots across the HaloTag protein spoted biosensor surface for both (FIG. 50A) Antibody #1 and (FIG. 50B) Antibody #2 injections are reported comparing values measured before and after sensor regeneration with the acidic 10 mM Glycine-HCl (pH = 2.4) regeneration buffer.

[0086] FIG. 51 depicts Antibody #1 off-rate pre- and post- acid buffer regeneration. Dot plot of off-rate kinetic values (Ad) obtained after Antibody #1 was injected over the sensor and allowed to dissociate for 30-minutes before (blue) and after (red) acid regeneration with 10 mM Glycine-HCl (pH = 2.4). Each dot represents the Ad measured at an individual HaloTag ligand on the SPOC biosensor, with the mean (dashed line) and standard deviation (error bars) overlayed (n = 197, **** = p < 0.0001).

[0087] DETAILED DESCRIPTION

[0088] Disclosed are compositions, devices, systems, and methods for performing real-time biosensing to detect and characterize molecular or biomolecule binding interactions, particularly for peptides, proteins, antibodies, drug molecules or candidates, or biologic drug molecules. In aspects provided, the disclosed compositions and methods perform real-time biosensing to detect and characterize molecule or biomolecule binding interactions with a high dissociation rate. In the disclosed aspects, the real-time biosensing is surface plasmon resonance (SPR) detection, more particularly real-time sensor-integrated proteome on chip (SPOC®) SPR-based detection.

[0089] In aspects of the disclosure, the compositions, devices, systems, and methods for used for synthesizing peptides, proteins, antibodies, drug molecules or candidates, and biologies (biologic drug molecules), to provide on-chip protein or drug or biologies libraries, for drug discovery' screening or diagnostics purposes. The compositions, devices, systems, and methods can be used for determining one or more kinetic binding parameter of the analyte to one or more biomolecule (e.g. peptides, proteins, antibodies, drug molecules or candidates, biologics / biologic drug molecules).

[0090] Surface plasmon resonance (SPR) is a real-time and label-free biosensing assay that can detect interaction of a molecule or biomolecule with an analyte or target in solution without the need for dyes (e.g. fluorescent dye) or other reporter tags (e.g., label-free). SPR bindingAtty. Dkt. No. 155529.00033

[0091] assays can directly measure the rates of association (ka) and dissociation (kd) of molecular interactions that can be used to calculate occupancy times, bound complex half-life (tl / 2), and equilibrium-dissociation constant (Kd). Sensor-integrated Proteome On Chip (SPOC®) is a next-generation protein biosensor technology that enables high density protein production directly onto SPR biosensors for cost-efficient and high-throughput real-time analyte screening. SPOC couples cost-efficient cell-free protein synthesis for high density on-chip protein libraries with label-free technologies like SPR, to improve real-time biomarker screening, kinetic evaluation of therapeutic biologies and drugs, and research into protein interaction networks.

[0092] As used in this disclosure, “association rate,’' “association rate constant'’ or “binding constant” (ka) refer to a rate or speed at which two molecules or substrates bind to each other, for example a protein and a ligand. “Dissociation rate” or “dissociation rate constant” (kd) refer to a rate or speed at which two molecules or substrates (e.g. a ligand dissociates from a protein). “Dissociation rate constant” (kd) is also referred to as equilibrium-dissociation constant.

[0093] One aspect is an on-chip library device of folded proteins, antibodies, drug variants, or biologic drug molecules configured to enable measurement of binding interactions for the plurality of folded proteins, antibodies, drug variants, or biologic drug molecules.

[0094] Another aspect of the disclosure is systems for performing real-time biosensing to detect and characterize molecular or biomolecule binding interactions, including surface plasmon resonance (SPR) detection, more particularly real-time sensor-integrated proteome on chip (SPOC®) SPR-based detection.

[0095] A further aspect of the disclosure is methods of detecting molecular or biomolecular interactions that have fast or rapid binding kinetics by performing SPOC binding assay for a biomolecule of interest. In some aspects the binding kinetics may be association rate and / or dissociation rate of the molecular interaction.

[0096] In some aspects, a method is provided for determining a dissociate rate of a biomolecular interaction comprising performing SPOC SPR binding assay for a biomolecule of interest.

[0097] In further aspects, a method is provided for monitoring transient off-target binding in biomolecular interactions comprising performing microarray assay or SPOC SPR binding assay for a biomolecule of interest.

[0098] In certain aspects, a method for detecting one or more analyte in a sample using a biomolecule array (interchangeable throughout the disclosure with microarray or nanowell array unless otherwise specified) is disclosed. The method can include exposing a sample to aAtty. Dkt. No. 155529.00033

[0099] biosensor surface (interchangeable throughout the disclosure with biosensor chip surface or second solid support substrate surface unless otherwise specified) of a biomolecule array. The biomolecular array can include a plurality of biomolecules coupled to the biosensor surface at discrete locations. The method can further include detecting one or more binding event involving one or more analyte in the sample to at least a portion of the biomolecules. The method can further include determining one or more kinetic parameter associated with the one or more binding event.

[0100] On-chip Library Devices

[0101] One aspect of the disclosure is an on-chip library device of folded proteins, drug molecules or biologic drug molecules that comprises: (i) a first solid support substrate surface comprising a plurality of microwells or nanowells. The microwells or nano wells comprise picoliter, nano-liter, or micro-liter volume reaction vessels; (ii) a second solid support substrate surface, with a plurality of capture moieties linked to the substrate surface. The second solid support substrate surface may constitute a glass surface, a chip surface, a slide surface, or a biosensor surface. The capture moieties comprise reagents or ligands that form covalent bonds with complementary peptide-tags or protein-tags, such as chloroalkane ligand that binds to haloalkane dehalogenase protein tag (HaloTag), SNAP-tag, CLIP-tag, ACP-tag, or Spytag-Spycatcher; (iii) a plurality of folded drug molecules or biologic drug molecules provided by an array of sealed microwells or nanowells containing isolated in vitro translation reaction mixture and a nucleic acid array on the first solid support substrate surface. The folded proteins, drug molecules, or biologic drug molecules are capable of being expressed as fusion proteins with peptide or protein tags and being captured via the capture moieties onto the second solid support substrate surface, via a fusion tag;

[0102] (iv) at least one of the folded proteins, drug molecules, or biologic drug molecules comprising one or more post translational modifications (PTMs).

[0103] The device is configured to enable parallel measurement of binding interactions for the plurality of folded proteins, drug molecules, or biologic drug molecules directly on the substrate without removal or transfer of the folded proteins, drug molecules, or biologic drug molecules (each also referred to herein as biomolecule) to a separate analytical platform.

[0104] In aspects, the drug molecules are antibodies or antigen-binding proteins.

[0105] In aspects, the fusion tag is a HaloTag, SNAP-tag, CLIP-tag, ACP-tag, or Spytag-Spy catcher.Atty. Dkt. No. 155529.00033

[0106] On-Chip Library Systems

[0107] Another aspect of the disclosure is a system of on-chip library of folded peptides, proteins, antibodies, drug molecules (or candidates), or biologic drug molecules. The system may comprise: (i) a first solid support substrate surface comprising a plurality of microwells or nanowells. The microwells or nanowells comprise pico-liter, nano-liter, or micro-liter volume reaction vessels; (ii) a second solid support substrate surface, with a plurality of capture moieties linked to the substrate surface. The second solid support substrate surface constitutes a glass surface, a chip surface, a slide surface, or a biosensor surface, the capture may moieties comprise antibodies, chemical linkers, affinity, agents, and / or a ligand for a fusion tag, such as a haloalkane dehalogenase tag (HaloTag) polypeptide (HaloTag ligand); (iii) a plurality of folded proteins, peptides, antibodies, drug candidates, or biologic drug molecules provided by an array of sealed microwells or nanowells containing isolated in vitro translation reaction mixtures and a nucleic acid array on the first solid support substrate surface. The folded peptides, proteins, antibodies, drug candidates, or biologic drug molecules are captured via the capture moieties onto the second solid support substrate surface via a fusion tag such as HaloTag protein. His tag, Flag peptide, or streptavidin tag.

[0108] In aspects, the on-chip library further comprises (iv) at least one of the folded peptides, proteins, antibodies, drug candidates, or biologic drug molecules comprising one or more post translational modifications (PTMs).

[0109] In embodiments, the folded peptides, proteins, antibodies, drug candidates, or biologic drug molecules to be synthesized on-chip are variants and / or clones derived from B cell NGS sequencing from immunized / inoculated animal models or subj ects / patients (humans), or discovered or derived from phase display libraries, yeast display libraries, mammalian display methods, RNA display methods, and / or ribosome display methods, in one or more affinity maturation cycles.

[0110] In embodiments, the folded peptides, proteins, antibodies, drug candidates, or biologic drug molecules to be synthesized on-chip are designed or predicted by artificial intelligence (Al) models, with or without use of target-specific training data. In some exemplary embodiments and methods, proteins or peptides or drug candidates to be synthesized on-chip are derived / discovered using a combination of the traditional approaches like B cells, phage or yeast display and Al driven design.

[0111] Disclosed are ‘‘microarray” or “nanowell array” that is an array for producing and screening functional proteins on biosensor surfaces that was previously developed to address the challenges of measuring proteomic interaction kinetics in high throughput (HTP). TheAtty. Dkt. No. 155529.00033

[0112] disclosed Sensor-Integrated Proteome On Chip (SPOC®) platform microarray or nanowell array involves in-silu cell-free protein expression in nano-liter volume (nanowells) directly from rapidly customizable arrays of plasmid DNA, facilitating simultaneous capturepurification of unique full-length folded proteins onto a 1.5 sq-cm surface of a single biosensor array or chip, such as a gold biosensor array or chip. A few to thousands of proteins can be captured on the biosensor array or chip. Arrayed SPOC sensors can then be screened by real¬ time label-free analysis, including surface plasmon resonance (SPR) to generate kinetic affinity, avidity data. Fluorescent and SPR assays were used to demonstrate zero crosstalk between protein spots. The recombinality of the SPOC protein array was validated by antibody binding assay, post-translational modification, mutation-mediated differential binding kinetics, and catalytic activity' screening on model SPOC protein arrays containing p53, Src, Jun, Fos, H1ST1H3A, and SARS-CoV-2 receptor binding domain (RBD) protein variants of interest, among others. Monoclonal antibodies were found to selectively bind their target proteins on the SPOC array. Further description of the microarray or nanowell array can be found in PCT Publication WO2025 / 102085, the contents of which are incorporated herein in their entirety'. In certain embodiments, the cell-free synthesis in nanowells can be extended to include synthetic or conjugated nucleotides or amino acids, with potential addition of natural or modified or synthetic enzymes and co-factors, to synthesize proteins carry ing unnatural amino acids to render new functionalities or enzymatic activities.

[0113] The microarray / nanowell array overcomes major challenges in resolving complexity in functional proteomics including (1) functionality in howto produce full length, folded proteins, (2) how to measure kinetic rates (affinities) of protein interactions with other proteins / biomolecules, how to make fully-folded functional proteins at an affordable cost, and scalability to investigate and detect 1000s of proteins and their interactions at once. The disclosed SPOC biosensor array or biomolecule array is an integrated platform where proteins are directly' expressed and immediately captured on the biosensor surface, eliminating the entire pipeline of recombinant protein production and purification. Capture molecules on the biosensor surface may include in situ cell-based or cell-free expressed or prior-expressed molecules directly spotted on the surface, such as antibodies, antibody fragments, nanobodies, peptides, scFvs, proteins, signaling molecules. In some embodiments, kinetic readouts occur through direct biomolecule sensing on the chip surface through technology' called surface plasmon resonance (SPR), providing real-time sensing of interactions between the molecules on the biosensor surface and the analytes in the sample. The simultaneous expression-captureAtty. Dkt. No. 155529.00033

[0114] method enables rapid alteration of the gene set on the chip surface, providing a customizable platform.

[0115] Antibody fragments come in a variety of engineered formats that differ in size, valency, and functional capabilities. The most common are Fab fragments, which contain one light chain and the VH-CH1 portion of the heavy chain, while F(ab’)2 fragments join two Fab units through hinge disulfides to create a bivalent molecule lacking the Fc region. The Fc fragment itself, composed of the CH2-CH3 domains, retains effector-function properties but not antigen binding. Smaller binding units include the Fv fragment (noncovalently associated VH and VL domains) and the widely used scFv, in which VH and VL are linked by a flexible peptide to form a compact monovalent binder. Variants such as disulfide-stabilized Fvs, diabodies, triabodies, and tetrabodies manipulate linker length and domain arrangement to create multivalent or multispecific binding constructs. Additionally, single-domain antibodies, including camelid VHH nanobodies, represent the smallest naturally occurring antigen-binding fragments. More complex architectures like bispecific fragments (example BiTEs, formed from two linked scFvs) provide dual-target engagement for advanced therapeutic or diagnostic applications. Together, these fragment classes enable tailored molecular designs for research, clinical use, and high-throughput screening workflows. Many of these antibody fragment constructs have one or more disulfide bonds, as post translational modifications, for proper folding, antigen binding and functionality.

[0116] Cells contain a diverse set of membrane-bound organelles that compartmentalize biochemical functions and maintain cellular organization, and associated membrane-bound proteins or intra-organelle proteins. The nucleus, enclosed by a double membrane, safeguards genetic material, while the endoplasmic reticulum and Golgi apparatus form a connected network for protein and lipid synthesis, folding, modification, and trafficking. Mitochondria, also double-membraned, generate ATP and regulate metabolic signaling, whereas lysosomes, peroxisomes, and endosomes mediate degradation, detoxification, and intracellular transport. Additional membrane-bound structures such as secretory and transport vesicles support cargo movement, and autophagosomes enable controlled recycling of cellular components. In plant and algal cells, chloroplasts and large vacuoles add further functional specialization.

[0117] SPOC biosensor arrays (biomolecule arrays) have applications described herein, ranging from immune response characterization (inflammatory / inhibitory markers, antibody reactivity and isotyping, etc.), immunogen discovery for vaccine development, pan-pathogen diagnostics / prognostics, drug and therapeutic discovery and characterization (including off-target binding detection and on-target validation), host-pathogen interactions (protein-protein,Atty. Dkt. No. 155529.00033

[0118] protein-small molecule, protein-antibody, protein-cell surface receptor, etc.), phenotype characterization (including mutation monitoring via altered kinetic binding), membrane protein expression via nanodiscs, companion diagnostic for clinical trials for pharmaceuticals and vaccines, target discovery for therapeutics / vaccines / drugs, among others. Though the primary purpose of this application is for biopharmaceutical applications, all applications herein described may be used in research use only, academic settings, or other non-clinical purposes, particularly in the context prior to or in the absence of regulatory approvals (such as CLIA, FDA, or other similar agency).

[0119] As discussed above, in various aspects, the systems can include a biomolecule array having a plurality of biomolecules or drug molecules coupled to a biosensor surface. In certain aspects, the biomolecules can be captured in an array format on any type of biosensor surface of a slide or component that is compatible with one or more biosensing techniques, such biosensing techniques capable of label-free and / or real-time biosensing. A non-limiting list of biosensing techniques capable of label-free and / or real-time biosensing via any number of approaches includes those that leverage biophysical, electrical, diffractometric and reflectometric methods such as SPR, RifS, SCORE, BLL focal molography, FET-based sensing, Raman-based sensing, plasmonic-sensing, or electrochemical -based sensing, without limitation. Protein microarrays are in part described in United States Patent No. US 10,983.118, which is incorporated by reference in its entirety. In certain embodiments, the terms “biomolecule array”, “captured protein array”, “captured antibody or drug library”, “biologies library'”, “on-chip library” have been used interchangeably.

[0120] In one aspect, the system can include a biomolecule array. The biomolecule array can include a first substrate and a biosensor surface coupled to the first substrate. A plurality' of biomolecules may be coupled to the biosensor surface. The biosensor surface can include an array of discrete locations, with each discrete location including a biomolecule of a plurality of biomolecules coupled to the biosensor surface. In certain aspects the biomolecules may be one or more of proteins or peptides or drug molecules or biologies, for drug discovery screening. In certain embodiments the biomolecules may be antibodies or antibody fragments. The system may comprise one or more additional substrate. For example, a second substrate may be connected to the first substrate of the biomolecule array. The plurality of biomolecules, or at least a portion of the plurality7of biomolecules, may be generated on the second substrate. The plurality of biomolecules or a portion of the plurality of biomolecules may be generated from a nucleic acid by in vitro transcription, in vitro translation, and in situ cell-free expression usingAtty. Dkt. No. 155529.00033

[0121] a cell-free lysate. The portion of the plurality of biomolecules may be purified. Further, the plurality of biomolecules may be captured to a biosensor surface of the biomolecule array.

[0122] In various aspects, the disclosed biomolecule array can include any type of substrate for supporting and / or providing a biosensor surface to which the biomolecules are coupled thereto. In certain aspects, the substrate can include clear material. In one or more aspects, the clear material can include fused silica. In certain aspects, the fused silica can include BK-7 fused silica, SF10 fused silica, or a combination thereof.

[0123] In various aspects, the biosensor surface of the biomolecule array may be a material different than the substrate. In various aspects, the biosensor surface can include any type of material compatible with label-free and / or real-time biosensing. In certain aspects, a nonlimiting list of example biosensor surface materials can include gold, silver, copper, aluminum, palladium, silicon, silicon dioxide, aluminum dioxide, and hafnium oxide. In one example aspect, the biosensor surface can include a thin layer of gold, e.g., a layer less than 200 nanometers (nm), less than 100 nm, or of from 1 nm-100 nm, 20 nm - 80 nm, or 45 nm - 55 nm. In various aspects, the biosensor surface may consist of a planar unstructured surface, or contain of a variety of microstructures and nanostructures, of which may consist of nanoparticles, nanorods, nanodisks, nanopores, and nanowires, among others.

[0124] In one or more embodiments, the biosensor surface may be coupled to the substrate using an adhesive layer. In certain embodiments, the adhesive layer can include any material capable of adhering the biosensor surface to the substrate and that is compatible with label-free and / or real-time biosensing. In various embodiments, the adhesive layer can include one or more of titanium, chromium, or tungsten.

[0125] In certain embodiments, the biosensor surface can be coated or passivated (e.g., a scaffold coating or passivation coating). For example, in one or more aspects, the biosensor surface can be coated with reagents, such as self-assembled monolayer (SAM) reagents, to passivate reactive gold surfaces of the biosensor surface and provide a scaffold for coupling antigens or other biomolecules of interest directly, or for coupling relevant ligands that are compatible with capture of the desired antigens or other biomolecules. For biosensor surfaces with gold coatings, as commonly used in SPR. SAM reagents may contain a thiol or silane moiety to absorb strongly onto the gold or gold-oxide surface and orient the SAM in a uniform manner, in certain aspects. In various aspects, the SAM may also include one or more fouling resistant surface chemistries (e.g., an anti-fouling coating) to reduce the non-specific binding of serum components to the biosensor surface, including but not limited to, N-hydroxy succinimide ester of 16-mercaptohexadecanoic acid (NHS-MHA), polyethylene glycolAtty. Dkt. No. 155529.00033

[0126] (PEG), polyethylene oxide (PEO), and other zwitterionic- or peptide-based formulations, among others. In other aspects the passivation layer also serves to couple with reactive groups present on biomolecules or ligands of biomolecules as means to capture biomolecules in array format and may consist of any but not limited to a plurality of 2-D or 3-D chemistries and hydrogels consisting of, for example, dextran, carboxymethlydextran, tetraethylene glycol, polycarboxylate, aliginate, and gelatin, among others.

[0127] In various aspects, the overall dimensions of the biomolecule array are sized to encompass the flow cell of the detector to be utilized, e.g., an SPR instrument or other detector. As discussed above, in various aspects, the biosensor arrays disclosed herein can include a plurality of biomolecules, including a plurality of mutationally scanned protein ligands, coupled to a biosensor surface, with each different biomolecule or ligand positioned at a discrete location on the biosensor surface. As used herein, a '‘discrete location” refers to a specific point or spot on the biomolecule array that is dedicated to the placement of a single biomolecule type (single mutationally scanned protein ligand) or mixture of biomol ecules. For example, the biomolecule array may include a discrete location (e.g., a spot), where there is only one biomolecule (e.g., antigen) bound to that discrete location.

[0128] In certain aspects, the biomolecule array can include a plurality of biomolecules with any number of and / or ty pes of biomolecules coupled to the biosensor surface. In one or more aspects, the biomolecule array can include 50 to 50,000 biomolecules or types thereof, 50 to 10.000 biomolecules or types thereof, or 100 to 5000 biomolecules or types thereof coupled to the biosensor surface. In various aspects, the biomolecule array can include at least: 2, 5, 10, 20, 40, 80, 100, 120, 140, 160, 180, 200, 220, 240, 260, 280, 300, 320, 340, 360, 380, 400, 420, 440, 460, 480, 500, 520, 540, 560, 580, 600, 620, 640, 660, 680, 700, 720, 740, 760, 780, 800, 820, 840, 860, 880, 900, 920. 940, 960, 980, 1000, 1500. 2000, 3000, 4000. 5000, 10,000, 30,000, 40,000, or 50,000 arrayed biomolecules or types thereof (e.g., mutationally scanned protein ligands). In various aspects, the range of spots (of biomolecules) per unit of area can range considerably. For example, in certain aspects, the spotting can range from five spots per square centimeter to 50,000 spots per square centimeter. In one example aspect, the spotting can be approximately 500 spots of proteins per square centimeter.

[0129] In various aspects, the biomolecule array can include any type of biomolecule coupled to the biosensor surface. In aspects, the biomolecules can include any molecule produced, used, or found in a biological system, including both naturally occurring biomolecules and synthetically-produced biomolecules. In various aspects, the biomolecules may derive from any animal or model organism including any mammal, such as a human, or non-mammal orAtty. Dkt. No. 155529.00033

[0130] single-cell organism or virus. In the various disclosed aspects, the biomolecule may comprise one or more of nucleic acid, peptide, polypeptide, protein, protein domain, phospholipid, immunoglobulin-like protein, small molecule, micelle, liposome, nanoparticle, polymersome, mesoporous particle, dendrimer, and aptamer, or a combination thereof.

[0131] In one or more aspects, the biomolecules can include one or more biomolecules associated with one or more clinical feature or outcome, such as one or more autoimmune or related disease, or other disease. For instance, in certain aspects, biomolecules can include experimental or long-standing disease-relevant antigens that can be pathologically targeted by the immune system of patients with autoimmune diseases and / or from various SARDs. In certain aspects, the biomolecules can include one or more antigens, e.g., for detecting one or more antibodies in a sample, and / or one or more analyte detection molecules for detecting one or more analytes in a sample. Analytes may include cytokines, chemokines, proteins, peptides, protein complexes, serum factors, small molecule chemicals, nucleic acids including aptamers, and whole cells, among others.

[0132] The terms "disease" or "‘disorder'’ or “condition” may be used interchangeably throughout this disclosure, and refer to a state of health of a mammal, for example a human, including any deviation from the normal state health of a mammal. The terms “patient” or “subject” may be used interchangeably in this disclosure and refer to all members of the animal kingdom prone to or suffering from a disease, disorder, or condition. The subject may be a mammal, e.g., a human or non-human.

[0133] In certain aspects, biomolecules can include nucleic acids, e.g., DNA, RNA, ssDNA, dsDNA, and / or ssRNA, lipids, phospholipids, cyclic citrullinated peptides (CCP), peptides, protein domains, full length proteins, or a combination thereof. In various aspects, one or more of the plurality of biomolecules can include one or more modifications and / or post-translational modifications. In such aspects, the modifications and / or post-translational modifications can include, but are not limited to glycosylation, phosphorylation, methylation, citrullination, sumolyation, disulfide bond formation, acety lation, lipidation, methylation, ubiquitination, S-nitrosylation. hydroxylation, amide formation, sulfation, myristoylation, deamidation, palmitoylation, carboxylation, formylation, O-linked glycosylation, S-palmitoylation, N-acetylation, N-myristoylation, cysteine modifications, or other modifications or one or more detection and / or affinity, tags or fusion proteins or peptides. In one aspect, a non-limiting list of antigens associated with one or more SARDs includes Smith, DNA, Ro / SSA, La / SSB, cNIA, Jo-1, PL-7, M-2, Ku, Scl-70, CENPB, Th / To, U1RNP. TRIM21. and ACPA. In various aspects, the analyte detection molecules (either printed onto the biosensor or captured in situ)Atty. Dkt. No. 155529.00033

[0134] can be any type of molecule capable of binding to one or more analytes in a sample. In certain aspects the analytes can include but are not limited to one or more cytokines, chemokines, autoantibodies, proteins, peptides, protein complexes, serum factors, small molecule chemicals, nucleic acids including aptamers, and whole cells or other relevant biomolecule, or a combination thereof. In certain aspects, the analyte detection molecules can include antibodies, single chain variable regions (scFv), nanobodies, antibody mimetics, or relevant analyte binding partners or a combination thereof. In one or more aspects, the analyte detection molecules are configured to bind to one or more analytes associated with the immune system, including but not limited to cytokines. In certain aspects, one or more of the plurality7of biomolecules may be prepared from one or more biological sources, e.g., recombinant expression in various bacterial, yeast, and / or mammalian cell lines. In the same or alternative aspects, one or more of the biomolecules may be prepared from non-recombinant native preparations, e.g., chromatin, intact snRNP, and nucleic acid extracts (e.g., from Calf Thymus and / or Salmon Sperm extracts). In certain aspects, one or more of the plurality7of biomolecules can be present in a cellular lysate, or lysate fraction, and coupled to the biosensor surface.

[0135] Proteins undergo a wide variety of post-translational modifications (PTMs) that expand their structural and functional diversity7beyond what is encoded by the genome. Phosphorylation is one of the most prevalent PTMs, involving the reversible addition of phosphate groups, usually to serine, threonine, or tyrosine residues, to regulate signaling cascades, enzyme activity, and protein-protein interactions. Glycosylation, occurring as N-linked or O-linked glycan attachment, affects protein folding, stability, trafficking, and immune recognition, especially for secreted and membrane proteins. Ubiquitination and SUMOylation involve the covalent attachment of small proteins (ubiquitin or SUMO) to lysine residues, modulating protein turnover, localization, and stress responses. Acetylation and methylation, commonly found on lysine or arginine side chains, play central roles in chromatin regulation and transcriptional control. Additional modifications such as lipidation (e.g., prenylation, palmitoylation), disulfide bond formation, and proteolytic cleavage influence membrane association, structural stability, and activation of protein precursors. Hydroxylation (e.g., proline or lysine hydroxylation in collagen) enhances structural rigidity and oxygen sensing. Nitrosylation and sulfenylation modulate cysteine reactivity in redox signaling networks. ADP-ribosylation, catalyzed by PARPs and related enzy mes, participates in DNA damage responses and chromatin remodeling. Carboxylation (e.g., in blood-clotting factors) enables calcium binding, while glycation represents a non-enzymatic modification linked to metabolic stress and aging. Together, these PTMs allow proteins to dynamically adapt to cellular state,Atty. Dkt. No. 155529.00033

[0136] environmental cues, and regulatory signals, creating a multilayered control system that shapes nearly every aspect of proteome function.

[0137] Enzymes can be broadly categorized into several major classes based on the type of chemical transformation they catalyze. Oxidoreductases mediate electron-transfer reactions, including oxidation, reduction, and redox balancing, and they are essential in metabolism, respiration, detoxification, and PTM formation such as hydroxylation or disulfide-bond formation. Transferases catalyze the transfer of functional groups, such as phosphate, methyl, glycosyl, or acyl groups, making them central to signaling (kinases), epigenetic regulation (methyltransferases, acetyltransferases), and protein maturation (glycosyltransferases). Hydrolases, one of the largest enzyme classes, cleave chemical bonds using water and include proteases, lipases, nucleases, and phosphatases, enabling protein turnover, nutrient breakdown, and reversible regulation of signaling pathways. Lyases catalyze the addition or removal of functional groups without hydrolysis or oxidation, often forming or breaking double bonds, while isomerases reorganize molecular structures, interconverting stereoisomers or structural isomers to support metabolism and biosynthesis. Finally, ligases (including synthetases) join molecules together using ATP or other energy sources, enabling DNA ligation, peptide bond formation, and assembly of complex biomolecules.

[0138] Each enzyme class contributes a distinct layer of biochemical capability7, collectively creating the dynamic, interconnected networks that sustain cellular life. Oxidoreductases shape redox homeostasis, transferases write and erase regulatory marks, hydrolases remodel and recycle biomolecules, and ligases construct macromolecular architectures. This functional diversity7underpins not only cellular metabolism and signaling but also biotechnological processes such as recombinant protein production, PTM engineering, and synthetic biology. Understanding how different enzyme families operate and how they interact is essential for designing efficient expression systems, tuning PTM fidelity, and building next-generation platforms like SPOC that precisely orchestrate thousands of biochemical reactions on-chip.

[0139] In various aspects, the biomolecules can be coupled to the biosensor surface using any techniques compatible for use in the systems and methods disclosed herein. In certain aspects, one or more of the plurality7of biomolecules can be prepared from one or more biological sources and either spotted and captured in array format from a crude lysate or from fractions containing a purified biomolecule. In various aspects, the biomolecules may be spotted and captured in array format using one or more of a plurality of contact and non-contact based microarray spotting approaches suitable for use in the systems and methods disclosed herein. In certain aspects, the microarray spotting approaches can be automated.Atty. Dkt. No. 155529.00033

[0140] In various aspects, the density of the biomolecules can be varied such as to overcome avidity effects which may mask important kinetic information of a given analyte-ligand interactions and / or lead to interactions which are too strong such that negligible dissociation is observed. Avidity effects occur when multivalent analytes are present in samples which can bind multiple ligands via any number of functional binding sites (not just a 1:1 interaction). In various aspects where avidity effects for multivalent analytes (e.g., immunoglobulins) are desired to be reduced and / or avoided, the density of the captured biomolecule ligands can be reduced such that the multivalent analyte is restricted to 1:1 interactions. This can be accomplished via a number of strategies including but not limited to vary ing the density of the capture surface chemistry (e.g., dextran, PEG, etc) to restrict the density of the capture biomolecule that can be achieved.

[0141] Additionally, if the capture biomolecule is mediated by a capture ligand (or capture tag) which is coupled to the biosensor surface (activated biosensor surface), as opposed to the biomolecule captured directly to the surface chemistry7itself, the concentration of the capture ligand can be adjusted such that there are fewer active areas on the sensor present for capture of biomolecules of interest thus leading to less densely captured biomolecules.

[0142] In certain aspects, functional protein or other biomolecules could be prepared via cell-free protein expression and coupled to the biomolecule array via microarray spotting techniques, or via in vitro transcription, in vitro translation and in situ capture-purification. Biomolecule functional protein, such as full-length folded protein, may be produced from nanowells. DNA (plasmid) encoding unique proteins of interest are printed or dispensed in nanowells. Full length proteins are expressed in situ from the DNA in nanowells in cell-free lysate. Expression in cell-free lysate, such as human cell lysate, ensures proper protein folding and protein functionality. Expressed proteins are capture-purified as an array of ‘pure protein monolayer spots’ on the biosensor surface. Functional protein production is customizable because only a plasmid with DNA sequence of interest is required to make custom protein.

[0143] By leveraging human cell-free in vitro transcription and translation (IVTT) lysate, a custom SPOC protein array production instrument is produced via in situ expression of full-length folded proteins, which are encoded by recombinant plasmid DNAs printed in nanowell slides. In certain embodiments the nanowell slide surface or the biosensor surface may comprise micro or nanostructures, pillars or trenches, micropores, meso pores or nanopores, that are produced using lithographic patterning and etching process or using chemical wet etching processes, in the wells and / or on the surface. In certain other embodiments, the nanowell slide surface or biosensor surface comprises coated films that have micro orAtty. Dkt. No. 155529.00033

[0144] nanostructures, pillars or trenches, micropores, meso pores or nanopores or similar other micro or nano paterns, in the wells and / or on the surface. Expressed proteins are simultaneously capture-purified as a monolayer of arrayed spots onto biosensor chips for label-free analysis by SPR. The entire process from protein array production on biosensor chip to label -free kinetic analy sis is called SPOC and is facilitated by the custom SPOC protein array production instrument. The SPOC protein array can produce multiple SPOC biosensors from a single nanowell slide. The SPOC protein array shown in FIG. 2Ahas four channels in each which can produce a nanowell slide in a semi-automated fashion; ultimately yielding up to sixteen SPOC chips in a single run.

[0145] An aspect is a system for integrated on-chip drug discovery' of peptide- or protein-based drug molecules or their conjugates, the system comprising:

[0146] a) a nanowell array comprising, a nanowell slide (first solid support substrate surface) comprising a plurality of wells, and a biosensor chip surface (biosensor surface or second solid support substrate surface). The biosensor chip surface comprises an array of discrete locations. A plurality of drug molecules is captured to the biosensor surface, to generate a plurality of on-chip drug molecules captured on the biosensor surface. Each discrete location comprises one drug molecule of the plurality of drug molecules captured to the biosensor surface. The plurality' of drug molecules may comprise one or more of peptide or protein. In embodiments, at least a portion of the plurality of drug molecules is generated on the biosensor surface from nucleic acid sequences using in vitro cell-free or cell-based expression system in the plurality of wells. In embodiments the plurality' of biomolecules is configured to bind covalently to the biosensor chip surface, which form a plurality7of on-chip drug molecules;

[0147] b) a fluidic module configured to deliver one or more analyte or target molecule to the plurality of on-chip drug molecules under controlled assay conditions. In some embodiments the plurality of on-chip drug molecules is configured to be regenerated after assaying with the one or more analyte or target molecule, for multiple successive assay and regeneration cycle; and

[0148] c) a detection or sensing module configured to measure binding interactions between the plurality of on-chip drug molecules and the one or more analyte or target molecule.

[0149] In some aspects of the disclosed systems, the plurality of drug molecules is a variant library' comprising variants of a single peptide or protein. In some aspects, the plurality7of drug molecules are antibodies or antigen-binding proteins.

[0150] In some aspects of the disclosed systems, each well of the plurality of wells of the nanowell array is at a temperature of about 10 deg. C to 37 deg. C.Atty. Dkt. No. 155529.00033

[0151] In some aspects of the disclosed systems, the nucleic acid sequences comprise 1) dual chain constructs comprising a heavy chain construct and a light chain construct and / or 2) single chain constructs at DNA concentration of about 0.1 nM to about 10 nM.

[0152] In some aspects of the disclosed systems, the nucleic acid sequences are added as a dual chain construct at a ratio of about 1:1 to about 1:3 molar ratio of heavy chain construct concentration to light chain construct concentration.

[0153] In some aspects of the disclosed systems, the binding interactions comprise mass spectrometer data, kinetic or equilibrium parameters comprising association rate, dissociation rate, equilibrium dissociation constant (KD), a maximum binding response, residence time, expression yield, binding free energy (AG), change in binding free energy (AAG), or combinations thereof, or combinations thereof

[0154] In aspects of the disclosed systems, the detection or sensing module is configured to generate interaction data, mass spectrometry data, or binding data comprising kinetic or equilibrium parameters, such as association rate, dissociation rate, equilibrium dissociation constant (KD), residence time, binding free energy (AG), change in binding free energy (AAG), or combinations thereof.

[0155] In aspects of the disclosed systems, the system is configured to perform synthesis or presentation of the drug molecules or variant library' and measurement of binding interactions on the same solid support surface without transferring the plurality of on-chip drug molecules off the biosensor chip.

[0156] In aspects of the disclosed systems, biosensor surface and the plurality of drug molecules are complexed with a detection tag or a ligand for the detection tag to allow the biosensor surface to capture the drug molecule via interaction of the detection tag and the ligand for the detection tag.

[0157] In aspects of the disclosed systems, detection tag and ligand for the detection tag belong to a system selected from the group consisting of a Halotag system, SNAP -tag system, CLIP-tag, ACP-tag, and Spytag-Spy catcher system.

[0158] In aspects of the disclosed systems, the detection tag or the ligand for the detection tag is configured to bind to the biosensor surface covalently.

[0159] In the disclosed aspects, the system further comprises a computational analysis module operatively coupled to the detection or sensing module and configured to transform the interaction data into drug-discovery outputs comprising at least one of affinity ranking, epitope binning competitive binding relationship, epitope classification, epitope mapping,Atty. Dkt. No. 155529.00033

[0160] polyreactivity' screening, off-target screening, aggregation assessment, thermal stability, or pH stability.

[0161] In aspects of the disclosed systems, the system is configured to assess aggregation propensity of the plurality of on-chip drug molecules by detecting changes in binding signal, kinetics, and / or surface behavior.

[0162] Another aspect is a system for integrated on-chip drug discovery of peptide- or proteinbased drug molecules or their conjugates, the system comprising:

[0163] (a) a solid support surface or a chip, comprising a plurality of discrete locations; (b) a biomolecular synthesis or presentation module configured to synthesize or present, on the solid support surface, a drug molecule variant library comprising plurality of protein, or peptide variants at each of the plurality of discrete locations,

[0164] wherein the biomolecular synthesis module comprises a nanowell array, configured to synthesize proteins or peptides or drug molecules from respective DNA templates in cell-free systems or cell-based systems in sealed nanowells, or nano-liter or micro-liter or pico-liter volume wells;

[0165] wherein the drug molecules are configured to bind covalently to the solid support surface or chip surface;

[0166] (c) a fluidic module configured to deliver one or more analyte or target molecule plurality of on-chip drug molecules on the solid support surface, under controlled assay conditions, wherein, the drug molecule variant library optionally is configured to be regenerated after assaying with each analyte or target, for multiple successive assay and regeneration cycles; and

[0167] (d) a detection module configured to measure binding interactions between the drug molecules and the one or more analytes or target molecules on the solid support surface, and to generate interaction data or binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or combinations thereof;

[0168] wherein the system is configured to perform synthesis or presentation of the drug molecules and measurement of binding interactions on the same solid support surface without transferring the drug molecules off the chip.

[0169] In some aspects, the plurality of drug molecules is a variant library comprising variants of a single peptide or protein. In some aspects, the plurality of drug molecules are antibodies or antigen-binding proteins.Atty. Dkt. No. 155529.00033

[0170] A further aspect is a system for integrated on-chip discovery and optimization of peptide- or protein-based drug molecules, or their conjugates, comprising:

[0171] (a) a solid support surface or chip, comprising a plurality of discrete locations;

[0172] (b) a biomolecular synthesis module configured to generate, synthesize, or present a plurality7of distinct protein- or peptide-based drug molecules at the discrete locations. In some aspects, the biomolecular synthesis module comprises a nanowell array configured to synthesize antibodies or proteins or peptides from respective DNA templates in cell-free systems or cell-based systems in sealed nanowells, or nano-liter, micro-liter, or pico-liter volume wells. In some aspects, the drug molecules are configured to bind covalently to the solid support surface or chip surface (on-chip drug variant library);

[0173] (c) a target interaction module configured to expose the drug molecules to one or more target molecules or analytes under controlled assay conditions. In some embodiments, the on-chip drug variant library7is configured to be regenerated after assaying with each analyte or target, for multiple successive assay and regeneration cycles; and

[0174] (d) a sensing module configured to acquire interaction data from each addressable site, the interaction data comprising binding data, kinetic parameters, equilibrium parameters, or a combination of these. The system is configured to execute generation of variants, interaction measurement, and computational analysis as a continuous on-chip workflow without physical transfer of the drug molecules or drug variant library between distinct experimental platforms.

[0175] In the aspects described above, the system comprises a data processing module configured to analyze the binding data to perform one or more integrated drug discovery functions selected from affinity, ranking, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping, on-target and off-target screening, poly reactivity screening, thermal stability analysis, aggregation assessment, hydrophobicity or hydrophilicity evaluation, solubility evaluation, pH stability evaluation.

[0176] In the aspects described above, the system comprises a computational analysis module operatively coupled to the detection or sensing module and configured to transform the interaction data into drug-discovery outputs comprising at least one of affinity, rankings, competitive binding relationships, epitope assignments, epitope maps, or residue-level energetic metrics, polyreactivity or off-target liabilities, aggregation or thermal stability or pH stress liabilities.

[0177] In the systems described above, the biomolecular synthesis module is configured to synthesize drug molecules under conditions optimal to minimize aggregation of drug molecules, as defined by synthesis temperature, time, reagents, co-factors, cell-free systemsAtty. Dkt. No. 155529.00033

[0178] In the systems described above, the proteins or drug molecules carry disulfide bonds and / or other PTM modifications.

[0179] In the systems described above, each of the discrete locations has a unique protein or drug variant molecule.

[0180] In the aspects described above, the computational analysis module is configured to perform affinity ranking of on-chip antibody, protein, or peptide variant library based on one or more binding parameters selected from association rate, dissociation rate, equilibrium dissociation constant (KD), residence time, binding free energy (AG), or change in binding free energy (A AG).

[0181] In the aspects described above, the system is configured to perform epitope binning by measuring competitive or non-competitive binding interactions by screening the on-chip antibody, protein, or peptide variant library (drug variant library) with the target followed by a purified variant or antibody, followed by regeneration and repeat cycles with the target and a different purified variant or antibody.

[0182] In the aspects described above, the computational analysis module is configured to perform epitope classification by grouping on-chip drug molecules based on shared or distinct binding behaviors observed across one or more target variants, domains, or reference binders.

[0183] In the aspects described above, the system is configured to perform epitope mapping by measuring binding interactions between on-chip antibody, protein, or peptide variant library and a plurality of target variants comprising single-amino-acid mutationally scanned variants of the target, or tiled multi-amino-acid mutationally scanned variants of the target, truncations, fragments, domains, or engineered target constructs.

[0184] In the aspects described above, the plurality of target variants differ by single amino acid substitutions, enabling single-amino-acid-resolution epitope mapping.

[0185] In the aspects described above, the system is configured to perform off-target screening by interrogating on-chip antibody, protein, or peptide variant library against one or more nontarget proteins, protein families, homologs, or proteome-derived components.

[0186] In the aspects described above, the system is configured to assess polyreactivity by measuring binding interactions of on-chip antibody, protein, or peptide variant library with heterogeneous molecular mixtures comprising serum, plasma, cell lysates, or panels of unrelated proteins, glyco-proteins, lipo-proteins, lipids, glycans, metabolites, PTM modified proteins, cell surface proteins.Atty. Dkt. No. 155529.00033

[0187] In the aspects described above, the system is configured to perform thermal stability analysis by subjecting on-chip drug molecule variant library to elevated temperatures prior to or during binding measurements and assessing changes in binding behavior.

[0188] In the aspects described above, the system is configured to assess aggregation propensity7of on-chip antibody, protein, or peptide variant library7by detecting changes in binding signal, kinetics, or surface behavior, or detecting a differently tagged variant coexpressed in the same nanowells, indicative of aggregation under defined assay conditions.

[0189] In the aspects described above, the system is configured to evaluate hydrophobicity or hydrophilicity7of on-chip antibody, protein, or peptide variant library7by measuring nonspecific binding, surface interactions, or binding responses to hydrophobic or hydrophilic reference surfaces or probes.

[0190] In the aspects described above, the system is configured to evaluate solubility7of on-chip antibody, protein, or peptide variant library by assessing binding performance under conditions that promote precipitation, surface fouling, or reduced functional availability7.

[0191] In the aspects described above, the system is configured to evaluate pH stability by measuring binding interactions of on-chip antibody, protein, or peptide variant library under acidic, neutral, and basic buffer conditions.

[0192] In the aspects described above, the computational analysis module is configured to use the plurality of binding parameters as input features, output labels, or both for training one or more machine-learning or Al models that predict binding behavior of antibody, protein, or peptide variants. In some aspects, AG values are computed from measured KD values under controlled assay conditions, and wherein AAG values are computed by comparing AG values of variant molecules to a reference molecule, and wherein the AG and AAG values are provided as quantitative training labels to the machine-learning or Al models. In some aspects, the computational analysis module is further configured to train the machine-learning models using multi-parameter binding profiles such that the models leam non-linear or non-additive relationships between association rate, dissociation rate, KD, residence time, AG, and AAG.

[0193] In some aspects, the machine-learning models are configured to predict at least one of a binding affinity, a kinetic rate parameter, a residence time, or a residue-specific energetic contribution for an unmeasured antibody, protein, or peptide variant.

[0194] In some aspects, the computational analysis module is further configured to iteratively retrain the machine-learning or Al models using newly generated binding parameters obtained from additional on-chip measurements of antibody, protein, or peptide variants.Atty. Dkt. No. 155529.00033

[0195] In some aspects, the plurality of binding parameters is generated under substantially identical assay conditions across the variants, enabling direct comparison and joint use of the parameters for machine-learning training.

[0196] In some aspects, the computational analysis module is configured to perform affinity maturation by identifying amino acid substitutions, insertions, deletions, or combinations thereof in antibody, protein, or peptide variants that are predicted to improve binding affinity to a target molecule.

[0197] In some aspects, the affinity maturation comprises generating a plurality of sequence variants that differ at one or more amino acid positions within complementarity-determining regions (CDRs), framework regions, or other binding-relevant domains.

[0198] In aspects described above, the sensing module is configured to measure binding interactions for the plurality of sequence variants and to generate kinetic or equilibrium parameters comprising at least one of association rate, dissociation rate, equilibrium dissociation constant (KD), residence time, binding free energy (AG), or change in binding free energy (A AG).

[0199] In some aspects described above, the computational analysis module is configured to train or update one or more machine-learning models using the measured binding parameters as training data for predicting affinity improvements associated with sequence modifications.

[0200] In some aspects described above, the machine-learning models are configured to jointly process sequence features and binding parameters to rank, select, or design affinity -matured antibody, protein, or peptide variants.

[0201] In aspects of the system as described above, the system is configured to execute an iterative affinity maturation loop comprising: (a) generating or selecting sequence variants using the machine-learning models; (b) measuring binding interactions of the variants on the solid support surface; and (c) retraining or refining the machine-learning models using newly generated binding data.

[0202] In some aspects, the affinity maturation is performed under multiple assay conditions. In some aspects, binding parameters obtained under the multiple assay conditions are jointly used for machine-learning training or prediction.

[0203] In certain embodiments of the system, the detection or sensing module measures the binding interactions to produce binding outcome data. The binding outcome data may comprise two or more of: association rate (ka), dissociation rate (kd), affinity constant (KD), residence time (tl / 2), equilibrium response level, binding stoichiometry, avidity signature, or competitive displacement metrics. In certain embodiments of the system, for each drugAtty. Dkt. No. 155529.00033

[0204] molecule at a discrete location, which may represent a synthesized variant of a variant peptide library, a linked digital record comprising: (i) a primary’ amino acid sequence, (ii) a structural representation, and (iii) one or more binding outcomes measured on the biosensor may be prepared. The linked digital record allows for sequence-structure-function correlations to be made. In certain embodiments of the system, the structural representation comprises one or more of: predicted 3D coordinates, backbone torsion angles, secondary’ structure annotations, solvent accessible surface area, electrostatic surface features, or paratope / epitope feature descriptors. In certain embodiments of the system, the structural representation may be generated by applying a structure prediction model to each sequence and the confidence metrics and / or per residue uncertainty' may be stored with the structural representation. In certain embodiments of the system, a mapping between sequence features and binding outcomes may be determined by computing one or more correlations between (a) residue level or motif level sequence features and (b) kinetic parameters including ka, kd, KD, or residence time. In certain embodiments of the system, the system computes “hotspot' residues or motif contributions by¬ perturbation analysis comprising one or more of: single residue substitutions, alanine scanning, saturation mutagenesis, or combinatorial scanning, and associates the perturbations with changes in kd, KD, and / or residence time based on analysis of variant peptide or protein drug molecules to which the system is applied. In certain embodiments of the system, epitope binning or competitive binding classification is performed and links bin assignments to sequence and structural features to derive epitope class-specific sequence-structure-function correlations. In certain embodiments of the system, single amino acid resolution epitope mapping is performed using variant peptide or protein drug molecules and the system stores, for each mapped epitope residue, a corresponding change in kd, KD, the AG and AAG values and / or competitive displacement, thereby correlating epitope residues to functional outcomes. In certain embodiments of the system, binding outcomes are measured against a plurality' of targets comprising at least a primary' target and one or more off targets, homologs, orthologs, or target molecule family members, and selectivity, correlations may be computed between sequence / structure features and differential KD and / or differential kd. In certain embodiments, the plurality of drug molecules or a plurality of target molecules includes species homologs used for preclinical model selection, and the system outputs a model selection score based on predicted or measured conservation of binding outcomes across species. In certain embodiments, developability measurements comprising one or more of: polyreactivity screening, non specific binding in serum / plasma / lysate, thermal stability challenge, pH stability challenge, aggregation propensity, solubility proxy, or hydrophobicity proxy, are generated andAtty. Dkt. No. 155529.00033

[0205] such measurements may be correlated with sequence and structural descriptors. In certain embodiments, multi objective rankings of variants or the plurality of the drug molecules are computed using a weighted combination of kd, KD, residence time, selectivity, and developability metrics, and the ranking is stored \with the corresponding sequence and structural representation. In certain embodiments, the system may be trained or fine tuned using a machine learning model using training examples comprising tuples of (sequence features, structural features, binding outcomes) measured on the biosensor. In certain embodiments, the machine learning model may comprise one or more of: a transformer, graph neural network, diffusion model, or ensemble model, configured to predict ka, kd, KD, residence time, or selectivity, from sequence and / or structure. In certain embodiments, the system performs active learning by selecting new sequences to synthesize based on uncertainty estimates of the machine learning model and / or expected improvement in one or more binding outcomes. In certain embodiments, the system supports a traditional hit to lead workflow by ingesting candidate sequences from one or more of: hybridoma derived antibodies, phage / yeast / mRNA display libraries, immunized animal repertoires, or rationally designed libraries, and generating sequence-structure-binding correlations used to guide lead optimization. In certain embodiments, the system supports Al based design by receiving candidate sequences output from a generative model and returning measured binding outcomes and correlation features as feedback to update the generative model. In certain embodiments, the system outputs a residue level attribution map indicating predicted or inferred contributions of residues to kd and / or KD, AG or AAG values, based on the measured binding outcomes and the stored structural representations.

[0206] In certain embodiments, the system generates, for each drug molecule or variant, a digital representation of the primary amino acid sequence and stores the representation in association with a unique spatial address on the solid support. In certain embodiments, the system further generates or receives a predicted structural representation of the drug molecule or variant derived from the primary amino acid sequence. In certain embodiments, the predicted structural representation comprises one or more of backbone conformation, secondary structure elements, surface topology,, solvent accessibility, electrostatic features, or predicted paratope geometry’. In certain embodiments, functional binding outcomes for each drug molecule or variant is measured using a biosensor. The functional binding outcomes may comprise two or more of association rate (ka), dissociation rate (kd), affinity constant (KD), residence time, AG or AAG values. In certain embodiments, the functional binding outcomes are correlated with both the primary sequence and the predicted structural representation to generate a sequence-Atty. Dkt. No. 155529.00033

[0207] structure-function relationship. In certain embodiments, the residue-level or motif-level contributions to binding outcomes may be computed by associating localized sequence or structural features with changes in kd, KD, or residence time, AG or AAG values across multiple variants. In certain embodiments, functional binding outcomes against a plurality of related targets comprising homologs, orthologs, or off-target proteins may be measured and selectivity metrics may be computed based on differential binding kinetics. In certain embodiments, selectivity metrics may be correlated with sequence and structural features to identify determinants of target specificity. In certain embodiments, one or more developability attributes selected from polyreactivity, non-specific binding, thermal stability, pH stability, aggregation propensity, or solubility proxies may be evaluated, and the attributes associated with sequence-structure-function relationships. In certain embodiments, a multi-objective score for each drug molecule or variant may be generated based on a weighted combination of binding kinetics, selectivity, and developability attributes. In certain embodiments, a machinelearning model may be trained or updated using training data comprising tuples of primary sequence features, structural features, and measured functional binding outcomes or energy considerations. In certain embodiments, the machine-learning model predicts one or more binding outcomes selected from ka, kd, KD, residence time, or selectivity. In certain embodiments, new drug molecules or variant sequences are selected for synthesis based on predicted improvement or uncertainty in the machine-learning model, thereby forming an active learning loop. In certain embodiments, candidate sequences may be generated by a computational or Al-based design model and may provide experimentally measured binding outcomes as feedback to refine the design model.

[0208] On-Chip Array Methods

[0209] Various aspects of the present disclosure relate to the production and expression of protein, including immunoglobulin (Ig)-like proteins, arrayed on biosensors, for example with an in vitro mix for in vitro transcription and translation (IVTT). An in vitro transcription and translation reaction as used herein refers to cell-free transcription and translation of RNA, mRNA, DNA, or other nucleic acids in plasmid (circular or linearized), genomic, fragments, or linear formats, as examples without limitation. In embodiments, a target nucleic acid is transcribed and translated for detection by a capture moiety. In embodiments, a capture moiety is transcribed and translated to detect a target. An “IVTT reaction yield’' refers to the amount of protein expressed in an IVTT reaction.Atty. Dkt. No. 155529.00033

[0210] The term “IVTT reaction mixture'’, “IVTT lysate”, “IVTT mix”, or ‘‘cell-free protein expression lysate” as used herein, refers to a solution comprising the reagents necessary to carry out an IVTT reaction. An IVTT reaction mixture typically contains a crude or partially-purified cell lysate, a suitable reaction buffer for promoting cell-free protein expression, and an RNA translation template, a DNA expression template, or both. Many IVTT reaction mixtures exist commercially or developed in research labs, including cell lysates derived from human (HeLa), mammalian (CHO, Chinese hamster ovary), bacteria (Escherichia coli. Agrobacterium tumefaciens), yeast (Saccharomyces cerevisiae), insect cells (Sf9), plants (wheat germ, Nicotiana benthamiana), as examples without limitation. This disclosure covers herein expression proteins in IVTT reactions with any and all IVTT reaction mixtures, including modified versions, derived from any source (any species, cell line, or other biological-based organism) without limitation. To facilitate disulfide bond formation, in some embodiments, redox altered IVTT reaction mixture can be used, such as E. coli Shuffle and Origami (Novagen) strains, including modified versions and derivatives thereof, though applicable to IVTT reaction mixtures from all species. Proteins may be expressed using IVTT reactions in tubes, plates, nanowells, or array formats, as examples without limitation.

[0211] In some aspects, the proteins or Ig-like proteins may be captured on the surface of a biosensor using a variety of methods. Ig-like proteins consist of variable and constant regions made up of heavy and light chains. These regions, or domains, can be rearranged in multiple formats to create unique Ig-like protein fragments. Human, camel (VHH), and shark (IgNAR) variations are depicted.

[0212] As an analyte is flowed in solution over the surface of a biosensor with proteins or Ig-like proteins captured on the surface, the analyte may interact with the proteins or Ig-like proteins constituting a binding event, which can be measured by a number of different technologies. Biosensor technologies for detecting biomolecular interactions include a plurality of label-free and real-time detection techniques, such as SPR, reflectometric interference spectroscopy (RifS), single-color reflectometry (SCORE), biolayer interferometry (BLI), focal molography, and field effect transistors (FET) based sensing, among others, without limitation.

[0213] In the disclosed methods, in certain aspects, binding events or interactions between Ig-like proteins may include kinetic, or binding affinity, measurements. Kinetic parameters may be measured, such as association constant, dissociation constant, affinity,, avidity, time of halflife. Binding events also constitute qualitative (presence / absence or yes / no) and quantitative data (any numerical, quantifiable output from the instrumentation or assay), which further define the binding interaction.Atty. Dkt. No. 155529.00033

[0214] In aspects, Ig-like proteins may include, as examples, single-domain antibody fragments (dAbs, or nanobodies), dual chain antibody fragments (Fabs), single-chain variable fragments (scFvs), camelid VH domains (VHH), shark VH domains (V-NAR, new antigen receptor), heavy-chain antibodies that lack light chains (HCAbs), cyclic scFvs, scFv region fused to a constant domain (eco), full-length antibodies (IgG, IgA, IgE, IgD, IgM and their subtypes or isotypes), bi-speciftc T-cell engager (BiTE), di-scFv, multispecifics (multispecific antibodies and fragments: dual-variable-domain Ig, tetravalent IgG-like antibodies, multispecific scFvs, IgG-scFv), diabody, minibody (scFv-CH3), scFab, scFv-zipper, scFv-Fc, triabody, tetrabody, IgN AR. bispecifics, trispecifics, camel Ig, monospecific, monovalent IgG, hdgG, dual affinity retargeting antibodies (DARTs), and bispecific killer cell engagers (BiKEs), Chimeric antigen receptors (CARs), CAR-T therapies, antibody-drug conjugates (ADCs), or any other combination of heavy chain, light chain, variable regions, and / or constant regions not listed, without limitation.

[0215] All Ig-like proteins are defined by two parameters: their binding epitope (interacting region on the target protein) and their affinity (strength of binding to the target protein). During antibody design, through artificial intelligence or other means, the target protein is defined and candidate antibodies are screened for affinity to the target and then epitope mapped to define the binding region. As such, thousands of antibody candidates may exist in the pipeline, requiring high-throughput assays for down-selection of the lead candidates. Affinity ranking based on the kinetic binding interaction is one aspect of lead candidate selection. In other aspects, the disclosed system and methods herein provide non-kinetic parameters of and screening tools for the candidate antibodies. Such parameters may include, without limitation: stability- at various temperatures, pH, buffers, salts, or other stabilizing media; expression or yield parameters based on screening different expression conditions to include, without limitation chaperones, disulfide bond catalysts, agitation, temperature, buffer, reducing or nonreducing conditions, enzymes: sequence parameters to include, without limitation signal peptides, stabilizing peptide sequences (positive versus negative charges, as an example), linker length and sequence, protease cleavage sites (or lack thereof), fusion tags (such as SUMO, as an example), chemical modification sites (or lack thereof). The platform described herein supplies information on antibody candidate characteristics listed above, provides a high-throughput sequence screening tool and method, and is applicable to all stages of the drug discovery- process from sequence design / discovery through stability and bioactivity evaluation for end-stage scale-up and manufacturing.Atty. Dkt. No. 155529.00033

[0216] As described above, proteins, such as Ig-like proteins, can be expressed with an in vitro mix for in vitro transcription and translation (IVTT) and in situ protein production. Expression can be achieved in an IVTT mix through transcription and translation of RNA, mRNA, DNA, or other nucleic acids in plasmid (circular or linearized), genomic, fragments, or linear formats, as examples without limitation. Many IVTT mix varieties exist commercially or developed in research labs, including cell lysates derived from human (HeLa), mammalian (CHO, Chinese hamster ovary), bacteria (Escherichia coli. Agrobacterium tumefaciens), yeast (Saccharomyces cerevisiae'), insect cells (Sf9), plants (wheat germ, Nicotiana henthamiana), as examples without limitation. This disclosure covers herein expression of Ig-like proteins with any and all IVTT mixes, including modified versions, derived from any source (any species, cell line, or other biological-based organism) without limitation. To facilitate disulfide bond formation, in some aspects, redox altered IVTT mix can be used, such as E. coh Shuffle82and Origami (Novagen) strains, including modified versions and derivatives thereof, though applicable to IVTT mixes from all species. Proteins may be expressed using IVTT methods in tubes, plates, nanowells, or array formats, as examples without limitation.

[0217] In various aspects, where biomolecules are not coupled directly to the surface of the slide, one or more capture or fusion tags can be employed which can be fused to biomolecules and proteins of interest (e.g., antigens or analyte capture molecules) and utilized to capture the proteins or other biomolecules in the desired array format on the biosensor surface. Capture ligands (also referred to as fusion ligands) can consist of a plurality of any number of ligands compatible with various tags (e g., capture or fusion tags) and may include but are not limited to Biotin (Streptavidin ligand), Glutathione (GST ligand), amylose (MBP ligand), Chloroalkane (HaloTag ligand), E-coil (K-coil ligand), K-coil (E-coil ligand), His tag or poly-His tag, FLAG peptide or FLAG tag, 06-benzylguanine (SNAP-tag ligand). 06-benzylcytosine (CLIP-tag ligand) SpyTag peptide (SpyCatcher ligand), SnoopTag peptide (SnoopCarther ligand), DogTag peptide (DogCather ligand) among others. Self-labeling tags, such as HaloTag, SNAP-tag, and CLIP-tag can be utilized to produce monolayers of covalently- captured protein or biomolecule spots on biosensor surfaces coupled with respective ligands, in addition to other covalent tags such as SpyTag, SnoopTag, and DogTag can be employed, and any other relevant high affinity interactions commonly used to couple ligand to biosensor surfaces which would be known to those familiar with the field (e.g., a plurality- of antibodybased capture via a plurality of epitope tags, E / K-coil interaction, the Streptavidin-biotin interaction, and GST / Anti-GST or Glutathione), in certain aspects. In embodiments, the folded protein, peptide, drug candidate, or biologic drug is fused with a peptide for covalentAtty. Dkt. No. 155529.00033

[0218] conjugation using enzy mes such as sortase or Sortase A (Sortase-Mediated Ligation) or site directed disulfide tethering using cystines or intein-mediated covalent ligation. Covalent capture modalities offer distinct advantage over other approaches, by enabling regeneration of the sensor surface during experimentation without loss of ligand from the biosensor surface when using acidic, basic, or high ionic strength regeneration buffers facilitating subsequent sample injections, in aspects. In certain embodiments, the present disclosure relates to systems and methods, compositions, and systems for covalent tagging, immobilization, conjugation, or assembly of biomolecules using irreversible or substantially irreversible chemical or enzymatic coupling reactions. Such covalent tagging chemistries may be employed to attach proteins, peptides, nucleic acids, complexes thereof, or other biomolecules to one another or to solid supports, including but not limited to planar surfaces, porous substrates, particles, microarrays, or biosensor interfaces.

[0219] In some embodiments, covalent tagging is achieved using enzyme mediated selflabeling systems, wherein a genetically encoded polypeptide tag selectively reacts with a complementary’ small-molecule or macromolecular or protein ligand to form a stable covalent bond, wherein the ligand may be immobilized to a substrate surface. In some embodiments, such systems may operate through nucleophilic substitution, acyl transfer, alkyl transfer, or related mechanisms and may proceed under aqueous, near-physiological conditions without the requirement for exogenous cofactors or activating reagents.

[0220] In certain embodiments, the covalent reaction is single turnover and results in permanent modification of the tagged biomolecule. In other embodiments, covalent tagging is achieved through spontaneous peptide-protein coupling systems derived from microbial structural or adhesin proteins. In such systems, a short peptide tag and a complementary protein domain associate and subsequently form an intermolecular isopeptide or amide bond between defined amino acid side chains. The bond formation may occur autocatalytically, without external enzymes, and yields a chemically stable linkage that is resistant to denaturation, hydrolysis, or mechanical stress.

[0221] In further embodiments, covalent tagging is accomplished using enzymatic ligation or transpeptidation reactions, including but not limited to reactions catalyzed by transpeptidases, ligases, or endopeptidases. Such enzymes may mediate site specific formation of peptide bonds between recognition motifs on target biomolecules and complementary nucleophilic groups on other biomolecules or surfaces (ligands immobilized on surfaces), enabling controlled covalent conjugation.Atty. Dkt. No. 155529.00033

[0222] In additional embodiments, covalent tagging is achieved using bioorthogonal chemical reactions between complementary reactive functional groups that do not substantially interfere with native biological processes. Such reactions may include cycloaddition reactions, condensation reactions, or other chemoselective coupling reactions, and may proceed in the absence of toxic catalysts or under conditions compatible with biological samples. In some embodiments, the reactive functional groups are introduced via genetic encoding, chemical modification, or incorporation of non-canonical amino acids.

[0223] In certain embodiments, covalent tagging is mediated by proximity-enabled or activation-dependent chemistries, wherein covalent bond formation occurs preferentially when two components are brought into close spatial proximity. Such systems may employ latent electrophiles, short-lived reactive intermediates, or photo-activated groups, thereby reducing non-specific conjugation and enhancing spatial or temporal control.

[0224] In some embodiments, the covalent tagging chemistries described herein are used individually or in combination, including orthogonal combinations that enable multiplexed or sequential conjugation events. The resulting covalent attachments may be used to facilitate stable immobilization, oriented capture, multivalent assembly, or repeated interrogation of biomolecules in analytical, diagnostic, synthetic biology, or high-throughput screening applications.

[0225] In certain embodiments, site-directed tethering is achieved by engineering one or more cysteine residues at defined positions on the fusion tag protein or peptide, thereby providing a uniquely reactive thiol group for selective covalent attachment to complementary thiol-reactive ligand on a solid support surface or biomolecule. The engineered cysteine may be positioned to control orientation, spacing, or accessibility of the tethered biomolecule. Covalent tethering may be carried out using thiol-reactive functional groups, including but not limited to maleimides, haloacetamides, vinyl sulfones, or disulfide-forming reagents attached / immobilized on a solid support surface, resulting in formation of stable thioether or disulfide linkages. In other embodiments, multiple cystines are engineered on the fusion tag protein or peptide that are designed and selected to form multiple disulfide covalent bonds with structurally matched (shape complementarity, like lock and key) complementary ligand peptide or protein bound on a solid surface. Such site-directed cysteine tethering enables precise and reproducible covalent immobilization of biomolecules on solid supports or to other biomolecules, while minimizing non-specific modification of native residues, and may be used alone or in combination with other covalent conjugation strategies described herein.Atty. Dkt. No. 155529.00033

[0226] In certain embodiments, covalent immobilization is achieved through incorporation of unnatural or non-canonical amino acids into a target biomolecule at one or more defined positions. Such non-canonical amino acids may bear reactive functional groups not commonly found in naturally occurring proteins, including but not limited to azides, alky nes, ketones, aldehydes, strained alkenes, strained alkynes, electrophiles, or photo-reactive moieties. The non-canonical amino acids may be introduced using genetic code expansion, engineered tRNA synthetase pairs, or chemical modification strategies. Following incorporation, the reactive functional groups may participate in chemoselective or bioorthogonal covalent reactions with complementary' groups presented on a solid support or a biomolecule, resulting in site-specific and irreversible attachment.

[0227] In certain exemplary embodiments, immobilization or capture on a solid support surface is achieved using ultra-high-affinity binding systems that rely on non-covalent interactions rather than covalent bond formation. Such systems may comprise complementary binding partners exhibiting equilibrium dissociation constants in few pico-molar or higher affinity, or in femto-molar or affinity range, or in atto-molar or affinity range, including but not limited to protein-protein, protein-peptide, protein-small molecule, or protein-nucleic acid interactions. The binding interactions may be driven by a combination of hydrogen bonding, electrostatic interactions, hydrophobic interactions, and shape complementarity7, and may exhibit extremely slow dissociation rates under assay conditions. In some embodiments, the resulting complexes are functionally irreversible over the time scale of analysis and provide stability comparable to covalent attachment while retaining the ability to be reversed under defined conditions. Such ultra-high-affinity non-covalent systems enable oriented, high-stability7immobilization and repeated interrogation of biomolecules on solid supports or w ithin analytical systems without need for chemical immobilization of biologics / proteins / drug / binder library to solid support surfaces (surface of chip or biosensor or slides).

[0228] In embodiments, folded proteins, peptides, drug candidates, or biologic drug molecules are captured covalently on the biosensor surface. In some embodiments, the drug molecules are antibodies or antigen-binding proteins. In embodiments, the biologics / proteins / drugs / binders can be immobilized on solid support surfaces using ultra high affinity^ but non-covalent tagging or immobilization methods.

[0229] HaloTag covalent binding chemistry7is based on an engineered haloalkane dehalogenase enzyme that reacts selectively and irreversibly with synthetic ligands bearing a chloroalkane or haloalkane moiety7. Halo proteins or its variants (HaloTag proteins) are fused to proteins / biologics of interest, to express fusion proteins in cell-free or cell-based systems,Atty. Dkt. No. 155529.00033

[0230] which then binds to solid support surfaces coated / immobilized with corresponding halo-ligand. The covalent binding chemistry is based on an engineered enzy me derived from a haloalkane dehalogenase scaffold that contains an active-site nucleophile positioned to react with chloroalkane or haloalkane-functionalized ligands bound to support surfaces. Upon binding of the ligand within the enzy me’s active site, the nucleophilic residue, typically a carboxylate side chain, attacks the carbon - halogen bond of the haloalkane moiety via a nucleophilic substitution mechanism, resulting in displacement of the halide leaving group and formation of a stable covalent ester linkage between the enzyme and the ligand. In some embodiments, the reaction proceeds efficiently under aqueous, near-physiological conditions and does not require additional cofactors or activating reagents. Because the engineered enzyme lacks the hydrolytic activity required to cleave the covalent intermediate, the resulting enzyme - ligand conjugate is effectively irreversible, yielding permanent attachment with high specificity and low background reactivity7. This mechanism enables controlled, site-specific covalent immobilization or labeling of fusion proteins containing the engineered enzyme domain on functionalized surfaces or within complex assay environments. The irreversible nature of the bond provides exceptional stability across a wide range of temperatures, buffers, and assay conditions, well suited for analytical, diagnostic, and high-throughput biochemical applications.

[0231] “Spytag-spyCatcher” as used herein refers to peptide - protein covalent coupling systems such as SpyTag - SpyCatcher based on spontaneous isopeptide bond formation between a short peptide tag and a complementary protein partner engineered from bacterial adhesin domains. Upon molecular recognition and complex formation, a reactive lysine residue on one binding partner forms a covalent amide (isopeptide) bond with an aspartate or asparagine residue on the complementary partner through an intramolecular acyl transfer reaction. This process is autocatalytic and proceeds without the need for external enzymes, cofactors, or chemical activation, occurring efficiently under aqueous, near-physiological conditions. The resulting isopeptide bond is chemically stable and resistant to denaturation, mechanical stress, and a wide range of buffer conditions, yielding irreversible and site-specific conjugation. Such peptide-protein covalent chemistry enables robust immobilization, capture, assembly, or stabilization of tagged biomolecules on solid supports, in solution, or within complex biological or analytical systems.

[0232] SNAP-tag covalent binding chemistry' is based on an engineered O6alkylguanine -DNA alkyltransferase derived protein that selectively reacts with ligands containing an O6alkylguanine or benzylguanine moiety. Upon binding of the ligand within the active site, aAtty. Dkt. No. 155529.00033

[0233] nucleophilic cysteine residue attacks the alkyl group of the O6alkylguanine substrate, resulting in transfer of the alkyl moiety to the cysteine side chain and formation of a stable thioether covalent bond. This reaction proceeds via a single-turnover, self-labeling mechanism in which the protein becomes irreversibly modified and inactivated following conjugation. The coupling occurs efficiently under aqueous, near-physiological conditions without the need for additional enzymes or cofactors and exhibits high specificity toward the engineered tag. The irreversible nature and chemical stability of the resulting linkage enable site-specific immobilization, labeling, or capture of SNAP-tagged biomolecules on functionalized surfaces or within complex assay environments.

[0234] CLIP-tag / ACP-tag systems use the E. coli acyl carrier protein (ACP) or a modified human 06-alkylguanine-DNA alkyltransferase (CLIP-tag). which react covalently with Coenzyme A (CoA) derivatives or 06-benzylcytosine (BC) derivatives, respectively.

[0235] In certain embodiments, covalent conjugation is achieved through selective reaction with amino acid side chains present on target biomolecules, including thiol groups of cysteine residues and primary’ amine groups of lysine residues or N-terminal amines. Cysteine-based conjugation may be carried out using thiol-reactive electrophiles such as maleimides, haloacetamides, vinyl sulfones, or related functional groups to form stable thioether linkages under mild aqueous conditions. Lysine-targeted conjugation may be achieved using activated esters, sulfonyl halides, isocyanates, or related amine-reactive chemistries to form covalent amide or urea bonds. Such reactions may be performed directly on native or engineered proteins, and in some embodiments are guided by steric accessibility, local microenvironment, or controlled reaction conditions to enhance selectivity’. These side-chain-directed covalent conjugation strategies enable robust attachment of biomolecules to other biomolecules or to solid supports, including surfaces, particles, or porous substrates, and may be used alone or in combination with other covalent tagging chemistries described herein.

[0236] In various aspects, where cell-free protein expression and in situ capture-purification of proteins is employed to form the desired biomolecule arrays, capture of non-IVTT (in vitro transcription translation) compatible or challenging to express components (e.g. nucleic-acid, phospholipids, CCP, and / or RF etc.) can be employed by one or more of a plurality of methods including but not limited to: (i) direct printing of the biomolecule into the nanowell slides for diffusion and capture onto the biosensor surface via appropriate tag and capture-ligand pairs during the cell-free protein expression incubation; or (ii) expression of tandem fusion tag protein domains whereby one tag (e.g.. HaloTag) is utilized to couple the tandem fusion protein to a ligand-coated biosensor surface while the remaining tag (e.g., SNAP -tag) is utilized toAtty. Dkt. No. 155529.00033

[0237] capture a compatible ligand-coupled antigen or biomolecule via incubation after the biomolecule array has been formed but prior to formal experimentation.

[0238] In certain aspects, post-translational modifications can be introduced post-production after the arrays have been formed on the biosensor surface via appropriate treatment with relevant enzy mes (e.g., PAD2) for creating modifications including but not limited to phosphorylation, glycosylation, citrullination, sumolyation. disulfide bond formation, acetylation, lipidation, methylation, ubiquitination, S-mtrosylation, hydroxylation, amide formation, sulfation, myristoylation, deamidation, palmitoylation, carboxylation, formylation, O-linked glycosylation, S-palmitoylation, N-acetylation, N-myristoylation, cysteine modifications, or other modifications.

[0239] In one or more aspects, whereas the above biomolecule arrays described involve capture of protein and formation of the arrays offline, biomolecule arrays as described prior can equally be prepared online of a biosensor instrument using one or more of a plurality of existing instrumentations such as the Carterra LSA.

[0240] As discussed above, in various aspects, the systems and methods can include a biomolecule array having a plurality of biomolecules or drug molecules coupled to a biosensor surface. In certain aspects, the disclosed biomolecule arrays are compatible with one or more biosensing techniques capable of label-free and / or real-time biosensing of one or more binding events between one or more of the biomolecules on the biomolecule array and one or more analytes. In various aspects, the biosensing techniques can utilize any type of detection capable of label-free and / or real-time biosensing described herein. In certain aspects, the detectors can include detectors utilized in one or more of SPR, RifS, SCORE. BLI, focal molography, FET- based sensing, Raman-based sensing, plasmonic-sensing, or electrochemical-based sensing among others.

[0241] In some aspects, Ig-like protein sequences may include signal peptides which target the protein to specific regions of the cell. Some of these compartments or regions may exist in IVTT mixes. In certain aspects, the Ig-like proteins may be internalized into these compartments during IVTT expression and production, thereby limiting their capture to the biosensor surface. As such, detergents or surfactants may be added, or excluded, from the IVTT mix to achieve optimal expression and production conditions which improve yield and capture onto the biosensor surface by disrupting the membrane components in the IVTT mix14 100. A non-limiting list of detergents and surfactants includes Triton X-100, Tween 20, N, N-Dimethylmyristylamine N-oxide, Nonidet P40, Polyethylene glycol sorbitan monolaurate, Poly oxy ethylenesorbitan monolaurate, CHAPS hydrate, ASB-14, C7BzO, PolyoxyethyleneAtty. Dkt. No. 155529.00033

[0242] (10) tridecyl ether (mixture of Cll to C14 iso-alkyl ethers with C13 iso-alkyl predominating), n-Dodecyl p-D-maltoside, Octyl P-D-glucopyranoside. Octyl p-D-1 -thioglucopyranoside, 3-(Decyldimethylammonio)-propane-sulfonate inner salt zwiterionic detergent, sodium dodecyl sulfate, and more. Detergents and surfactants may be added at any range of concentrations, including, without limitation, 0, 0.01, 0.05, 0.1, 0.25, 0.5, 1% and any number or decimal above or below those listed. Other denaturing agents may, or may not be, included to achieve a similar outcome by disrupting structures within the IVTT mix, which may or may not induce protein refolding and disruption of inclusion bodies or aggregation. As non-limiting examples: urea, dimethylsulfoxide, isoproponal, oxidized glutathione, and more. Denaturing agents may be added in a range of concentrations measured in different units, such as 0-5% inclusive, 0-4 M inclusive, or any number or decimal above or below listed values. As noted here, not all Ig-like proteins require detergents or surfactants, and therefore this is not an absolute requirement of the system or methods described within this disclosure.

[0243] In some aspects, the addition of expression cofactors may be required to achieve optimal expression. The inclusion or exclusion of such cofactors may be Ig-like proteinspecific and therefore vary for each protein arrayed on the biosensor or other substrate. In some aspects, the inclusion or addition of expression cofactors to the reaction mix can be achieved by direct expression via DNA (plasmid or linear) in the IVTT mix simultaneously expressing the Ig-like protein. To achieve chaperone screening from DNA expression within the IVTT mix, plasmid or linear DNA can be directly printed in an array or included within the tube or plate which contains the plasmid for the desired Ig-like protein. Alternatively, recombinant proteins can be added to the IVTT mix before, during, or after expression of the Ig-like protein. Optimal timing of the addition or expression of the cofactor may be required. In such cases, inducible expression via regulatory’ promoters can be used for directly expression cofactors, or timed addition can be an alternative method for control the cofactor inclusion in the IVTT mix. Cofactors may include chaperones which assist with protein folding. Such chaperones have been included in a variety of combinations, and some of the most commonly used chaperone combinations are available in plasmid formats commercially. Examples of chaperones that can be included are given here, without limitation: dnaK, dnaJ, grpE, groES, groEL, grpE, tig, trigger factor, sHsp family, HSP60 / HSP10 family, HSP70 family, HSP90 family, prefoldin family, co-chaperones Hsp40 family, co-chaperone NEF family, etc. Chaperones can be included (or excluded) alone or in combination in any variation, without limitation. Inclusion or exclusion of specific chaperones may increase yield, solubility, expression, capture, stability, and / or bioactivity by enabling proper folding conformations of the Ig-like proteins.Atty. Dkt. No. 155529.00033

[0244] As noted here, not all Ig-like proteins require chaperones, and therefore this is not an absolute requirement of the system or methods described within this disclosure. Once the optimal combination of chaperones is determined for any given Ig-like protein, this can then be applied to all downstream assays, production, and manufacturing.

[0245] Other materials may be added or excluded from IVTT mixtures to improve Ig-like protein folding, expression, and yield, for example: bacterial glycoengineering for PTMs required for full-length antibodies. Glycosylation enzymes may be added, or excluded from IVTT mixes to achieve addition or absence of N- or O-linked or other glycans; inclusion or exclusion of protease inhibitors in the IVTT mix or derive IVTT mix from cell lines or organisms lacking specific proteases or with mutated proteases; and inclusion or exclusion of nanodiscs or other lipoparticles to provide a membrane or hydrophobic region to facilitate proper protein folding, reduce aggregation, permit signal peptides to home properly, or other similar mechanism to improve expression, yield, and capture of Ig-like proteins onto the biosensor surface.

[0246] In some aspects, additional genetic constructs or recombinant proteins may be required to achieve optimal expression of Ig-like proteins for capture onto biosensors. These components may improve stability, yield, expression, folding, or reduce aggregation or any other mechanism to permit expression and capture of Ig-like proteins onto biosensor surfaces. Components may be added or excluded in any combination to achieve the expression and capture of Ig-like proteins onto biosensor surfaces. For example, sortase A and HRV3C protease may be included for cyclic scFv formation.

[0247] The amino acid sequences can be included or excluded from the Ig-like protein genetic constructs to achieve targeting by sortase A and HRV3C protease. Sortase A and HRV3C protease can be added to the IVTT mix either through inclusion of a genetic construct encoding sortase A or HRV3C protease or through the direct addition of recombinant sortase A or HRV3C, derived from any species or biologic source. Sortase A or HRV3C may also be added post-expression (post-translationally) by direct addition to the biosensor surface or at any point during the expression reaction. Sortase A and HRV3C may be added together or subsequently, encoded on the same or different genetic constructs, and used in any variation of combinations and timings to improve stability, yield, expression, folding, or reduce aggregation or any other mechanism to permit expression and capture of Ig-like proteins onto biosensor surfaces. A further example is co-expression of disulfide bond catalysts. A non-limiting list of examples of disulfide bond catalysts includes: Ervlp. DsbB, VKOR, DsbC, PDI, DsbA, DsbD. DsbG, SurA, FkpA, PpiA, PpiD, Skp, DegP, Erol.Atty. Dkt. No. 155529.00033

[0248] Disulfide bond catalysts may be included or excluded from the IVTT mix in a variety of ways. They may be added in any combination at a variety of concentrations via the following non-limiting examples: encoding by genetic constructs (one or more per construct), addition of recombinant proteins derived from any biological source. The disulfide bond catalysts may be added in a variety' of timings: included in the IVTT mix prior to or in conjunction with the Ig-like protein genetic construct, timed addition at any point during the expression process, or post-expression treatment of the IVTT mix or biosensor surface. The optimal ratio and combination of disulfide bond catalysts will be determined by those that improve stability, yield, expression, folding, or reduce aggregation or any other mechanism to permit expression and capture of Ig-like proteins onto biosensor surfaces.

[0249] Biopharmaceutical companies invest billions of dollars annually in developing new drugs to prevent and treat diseases, yet only about 10% of drug candidates that enter development pipelines and clinical trials ultimately receive regulatory approval. The majority of failures are attributed to inadequate efficacy and / or safety', which are often linked to the on-and off-target binding properties of drug molecules. As a result, there is a renewed emphasis on designing new molecular entities (NMEs), particularly biologies or biologic drugs, that exhibit high affinity, specificity, and selectivity to their intended targets in early preclinical phase, towards furthering the candidates with improved profiles along the development pipelines. Currently, deep characterization of binding kinetics is typically performed during the lead selection process, and safety assessments are often conducted during costly in vivo animal studies. To improve success rates, it is critical to assess kinetic binding and safety profiles earlier in the preclinical workflow ideally in the design phase while striving to maintain a broader diversity of lead candidates. Achieving this requires new or improved methods to accelerate the design, build, and test cycles, for iterative improvement and down selection of most optimal lead candidates. Such advancements could significantly enhance clinical success rates, saving biopharma hundreds of millions of dollars annually, and expediting the delivery of new and improved therapies to patients.

[0250] The class of antibody therapeutics and biologies (biologic drugs or molecules) has undergone significant innovation over the past two to three decades, driving the development of newer therapeutic modalities and yielding highly effective drugs, even as we enter the new age of Al-driven drug design. Following the success of full-length antibody therapeutics such as trastuzumab (Herceptin), bevacizumab (Avastin), and adalimumab (Humira) at the turn of the century, the therapeutic landscape has expanded dramatically to include a diverse and powerful array of novel single-chain antibody derivatives. These include single-chain antibodyAtty. Dkt. No. 155529.00033

[0251] formats such as single-chain variable fragments (scFvs), single variable domain of heavy-chain antibodies (VHH, also known as nanobodies, trademark of Ablynx N. V). and shark-derived variable new antigen receptors (VNARs); this is in addition to antibody mimetics such as designed ankyrin repeat proteins (DARPins), and a diverse class of Al designed protein binders. Single-chain antibodies, in particular, offer several advantages due to their smaller size, which often enhances tissue penetration and have fewer functional components outside of the antigen binding region compared to most mammalian antibodies, potentially lowering immunogenicity. These compact structures are also easier to produce using a range of expression systems, including yeast and bacterial platforms such as E. coli.

[0252] Nanobodies, which are derived from the antigen-binding region of specialized singledomain antibodies found in camelids (such as llamas, camels, and alpacas), hold immense promise for current and future clinical applications. These compact molecules, ranging from 12 to 15 kDa in size, consist of a single immunoglobulin domain that exhibits high-affinity binding (nM to pM range) even though it lacks a light chain. The small size and unique structure of nanobodies enable them to access cryptic antigens that conventional full-length antibodies cannot reach. Additionally, in certain modalities, their compact nature can facilitate cellular uptake into the cytoplasm, allowing them to target intracellular antigens. Like the variable regions of IgG molecules, nanobodies possess three complementarity-determining regions (CDRs) that form the antigen-binding site (paratope). However, unlike IgGs, nanobodies are more thermostable, less prone to aggregation, and exhibit a shorter half-life — though the latter can be extended through conjugation with specific proteins or fusion with albumin-binding VHH.

[0253] Given the therapeutic promise of antibody fragments and protein binders, drug developers are increasingly focusing on engineering these molecules to address previously encountered challenges related to affinity, specificity, selectivity, and tissue distribution observed in earlier generations of antibody drugs. These efforts aim to enable access to previously obscure targets or specific epitopes, significantly expanding therapeutic possibilities. The growing popularity of sc-antibody drug discovery’ is evident from the proliferation of commercially available synthetic libraries, including fully naive libraries, designed for rapid target screening and high throughput binder identification. Al-driven engineering of antibody fragments and protein binders offers the potential to further accelerate the development of new therapies by optimizing paratopes for specific clinical targets. However, realizing this potential requires the development of new wet lab techniques capable of high-throughput, deep characterization of these engineered molecules. Such techniques areAtty. Dkt. No. 155529.00033

[0254] essential for rapid testing, analysis, and iterative improvement cycles to identify leads with very high affinity (or affinity optimized for specific therapeutic modalities), high specificity, and high selectivity, ensuring their readiness for prechnical and clinical testing.

[0255] As part of design-build-test-leam (DBTL) cycles in drug development, affinity maturation campaigns are frequently undertaken to iteratively enhance the binding properties of identified binders and lead candidates. The initial step in this process involves deep characterization of the sc-antibody paratopes. This is typically achieved through comprehensive degenerate mutagenesis or computationally driven designer mutations introduced into the CDRs, linkers, and / or scaffold regions. The SPOC technology offers a transformative solution for high-throughput deep kinetic characterization of sc-antibody (e.g., scFv. Fab and VHH) variants, enabling the production of SPR biosensor chips with hundreds to thousands of unique sc-antibody drug candidates produced and captured on chip. The platform allows for direct and simultaneous high-resolution kinetic measurements of analyte binding (e.g., antigen targets in solution) across the entire on-chip scFv, Fab or VHH library in a single assay. This approach facilitates the production of thousands of properly folded proteins in discretely separated and isolated nano wells, within a 1.5-square-centimeter area, which are then directly capture-purified onto SPR biosensor chips. The SPOC system enables the characterization of binding interactions for up to 1000-2400 sc-antibody variants against their antigen target or alternative analytes of choice (e.g., potential off-targets).

[0256] Regeneration of the sensor surface

[0257] Disclosed are methods for screening drug candidates that include sequential, multiple screening assays with regeneration of nanoarray chips between each screening assay. The disclosed methods comprising regeneration of nanoarrays, biomolecules bound to a chip (on-chip biomolecules), drug molecules bound to a chip (on-chip drug molecules), that can occur between various screening methods.

[0258] Biomolecules and drug molecules, such as antibodies and antibody fragments are covalently captured on the SPR biosensor chip, allowing for enhanced stability and offering the potential for multiple rounds of regeneration and follow-on assays. This feature ensures that a single chip may be reused for collecting replicate data, further improving cost-efficiency and throughput in the characterization and validation processes. Importantly, the SPOC workflow only requires the DNA sequence library encoding the antibody fragments or protein binders, eliminating the need for expression and purification of proteins. By leveraging plasmid or linear DNA and cell-free expression systems to convert sc-antibody gene / DNA libraries into protein libraries, the SPOC platform dramatically reduces the time and cost associated withAtty. Dkt. No. 155529.00033

[0259] obtaining high affinity resolution SPR kinetic data which provides a wealth of information about the binding characteristics of each sc-antibody on the sensor, including (Rmax), on-rate (ka), off-rate (kd), affinity (KD), and half-life (tl / 2). Compared to traditional recombinant production approaches, this innovation accelerates the testing and validation process, making it feasible to generate the large-scale, high-quality wet lab data required to train Al models for better prediction accuracies.

[0260] Regeneration of the biosensor chip may include regeneration between sequential introductions of the target molecule and / or each target molecule variant by contacting the plurality of drug candidates with a regeneration reagent to remove bound target molecules.

[0261] In various aspects, regeneration of the sensor surface can be employed to facilitate repeated and serial probing of the biomolecules on the array with a plurality of analytes of interest from a plurality of biological sources of interest. In certain embodiments, systems and methods are provided for regeneration of an on-chip library of drug candidates immobilized on a solid support, such that the same library may be reused for multiple assay cy cles or for interrogation against multiple targets or analytes. Regeneration may be achieved by contacting the immobilized drug candidates with one or more regeneration reagents selected to disrupt binding between the drug candidates and a previously bound target or analyte without substantially affecting the integrity, conformation, or surface attachment of the drug candidates. Such regeneration reagents may include, but are not limited to, solutions of altered pH. high or low ionic strength buffers, chaotropic agents, detergents, competitive ligands, or other dissociation-promoting compositions. Following removal of the bound target or analyte, the on-chip library may be returned to assay-compatible conditions, thereby restoring the binding functionality of the immobilized drug candidates. This regeneration process enables repeated or sequential assays using the same on-chip library, reduces reagent consumption, and supports high-throughput or multiplexed evaluation of drug candidates across multiple binding experiments. Ideal regeneration injections would not lead to significant loss of biomolecule ligand capture level or integrity yet would effectively strip away bound analytes from previous injection facilitating another round of analyte binding without interference from previously bound analytes. In the examples, the inventors used Halo-tag to covalently capture proteins on solid support surfaces (biosensor chip) and regenerated the on-chip proteins multiple times, as many as 30+ regenerations.

[0262] Provided below are systems and methods that can be used for an integrated on-chip drug discovery workflow. The on-chip biologics / protein / binder library described herein can be used sequentially for multiple drug screening assays, for example by regeneration of the chipAtty. Dkt. No. 155529.00033

[0263] or biomolecules or drug candidates immobilized on the chip (e.g. on-chip library of drug candidates or biomolecules). This can be use of a single on-chip library for multiple assays with regeneration in between, use of multiple on-chip libraries with the same biologics / protein / binder library and created from same template DNA printed in nanowell arrays, each chip used for one or more assays, or a combination of the two. The systems and methods outlined below can be used alone or in combination. If the systems or methods are used in combination, they can be used or performed sequentially in any order.

[0264] On-chip library of covalently captured biologics / proteins screened with targets / analytes and ranked by their performance

[0265] In certain embodiments, systems and methods are provided for affinity ranking of drug variants, including but not limited to protein variants, antibody variants, peptide variants, or other binding molecules, based on their binding interactions with a target molecule or analyte. Such systems and methods may comprise generating a plurality of drug variants immobilized on chip, with variants differing in one or more sequence, structural, or chemical features, followed by measurement of binding interactions between each variant and the target or analyte. The measured data is used for ranking the variants according to one or more binding parameters. The binding parameters may include, but not limited to equilibrium dissociation constants (KD), association rate constants (ka), dissociation rate constants (kd), residence time, expression yield (R) or combinations thereof. Affinity ranking of on-chip drug or protein variants may be performed in a comparative or competitive format, under substantially identical assay conditions, to enable direct assessment of relative binding performance. In some embodiments, the ranked variants are selected, prioritized, or further optimized based on predefined affinity thresholds or multi-parameter criteria, thereby enabling systematic identification of variants with improved binding characteristics for therapeutic, diagnostic, or analytical applications.

[0266] Epitope binning or classification

[0267] In certain embodiments, systems and methods are provided for epitope binning or epitope classification of an on-chip library of drug candidates by sequentially screening each drug candidate as an analyte against the immobilized library. In such systems and methods, a plurality of drug candidates are immobilized on a solid support in a defined spatial arrangement, and individual drug candidates, or subsets thereof, are introduced sequentially in solution as analytes in the presence of a target molecule or in a competitive binding format. Binding responses are measured to determine whether the analyte drug candidate competes with, blocks, or is simultaneously compatible with binding to each of the on-chip immobilizedAtty. Dkt. No. 155529.00033

[0268] drug candidate library, thereby enabling assignment of the candidates into distinct epitope bins or overlapping binding groups. The sequential analyte screening may be performed with regeneration steps between cycles to remove bound components while preserving the immobilized library. This approach enables systematic, high-throughput epitope binning and interaction mapping of large drug libraries using a single on-chip format and supports prioritization of drug candidates based on epitope diversity’, competition profiles, or combinatorial binding behavior.

[0269] In certain embodiments, systems and methods are provided for indirect epitope mapping or classification of drug candidates by performing competitive binding assays with one or more reference antibodies, binders, or ligands having known binding epitopes on a target molecule. In such systems and methods, a drug candidate is assessed for its ability to compete with, inhibit, or displace a reference binder when both are exposed to the target under controlled conditions. Changes in binding signal, kinetics, or occupancy relative to the reference binder are used to infer whether the drug candidate binds to the same epitope, an overlapping epitope, or a spatially distinct epitope on the target. The competitive assays may be conducted in sequential, simultaneous, or order-of-addition formats and may be repeated across multiple reference binders to generate an epitope competition profile for the drug candidate. This indirect epitope mapping approach enables rapid classification and prioritization of drug candidates based on epitope location and overlap without requiring direct structural determination of the binding interface.

[0270] Embodiments include high resolution epitope mapping. In certain embodiments, high-resolution or single amino-acid resolution epitope mapping is achieved by sequentially screening on-chip library of drug variants against a panel of target molecule variants generated by deep mutational scanning. The target variants may comprise single amino-acid (AA) substitutions, or tiled multi-AA substitutions, including but not limited to alanine substitutions, introduced across one or more regions of the target, or whole of the target. In such systems and methods, each target variant is introduced sequentially as an analyte to interact with the immobilized drug variant library, with regeneration cycles between each target variant, and binding responses are measured under substantially identical assay conditions. A reduction or loss of binding signal of an on-chip drug variant relative to the corresponding wild-type target is indicative that the substituted amino acid residue is involved in, or in close proximity to, the binding epitope recognized by that specific drug variant. By correlating binding changes across the panel of target substitutions, residue-level epitope maps may be generated, enabling precise delineation of contact residues, epitope boundaries, and energetic hotspots. This approachAtty. Dkt. No. 155529.00033

[0271] supports systematic, high-throughput, and quantitative epitope mapping of drug candidates without requiring direct structural determination of the drug-target complex.

[0272] Binding Free Energy

[0273] In certain embodiments, the binding free energy (AG) of an antibody-target interaction is derived from experimentally measured equilibrium binding parameters obtained from surface-based kinetic assays. The binding free energy AG may be calculated as a function of the equilibrium dissociation constant (KD), absolute temperature, and a gas constant, such that AG is proportional to the natural logarithm of the equilibrium dissociation constant multiplied by a temperature-dependent scaling factor (AG equals RT multiplied by the natural logarithm of KD). In some implementations, AG is determined according to a relationship in which stronger binding affinities (lower KD values) correspond to more favorable, more negative free energy values, while weaker binding affinities correspond to less favorable, more positive free energy values.

[0274] In further embodiments, a change in binding free energy (AAG) associated with a specific amino acid substitution, modification, or perturbation is determined by comparing the binding free energy of a variant antibody sequence to that of a corresponding reference or parent antibody sequence. The AAG value may be calculated as the difference between the AG of the variant and the AG of the reference, such that positive AAG values indicate a reduction in binding favorability attributable to the residue change, and negative AAG values indicate an improvement in binding favorability. In some embodiments, AAG values are calculated independently for multiple single-residue variants, enabling assignment of a residue-specific energetic contribution (AG_residue) to individual amino acid positions within the antibody.

[0275] In certain embodiments, AG and AAG values are calculated using equilibrium binding constants derived from kinetic rate measurements, including association and dissociation rate constants, or directly from steady-state affinity measurements. The calculations may be performed under standardized assay conditions, including controlled temperature, buffer composition, and surface density, such that relative energetic contributions across variants are comparable. The resulting AG and AAG values may be stored, visualized, or computationally processed to generate residue-level energetic maps, which may be used to guide antibody engineering, affinity maturation, epitope mapping, or computational modeling of antibodytarget interactions.

[0276] In certain embodiments, the disclosed methods enable calculation of residue specific energetic contributions (AG residue ~ contribution to binding from each amino acid residue) to antibody target binding using single amino acid resolution binding data derived from surfaceAtty. Dkt. No. 155529.00033

[0277] based kinetic measurements such as surface plasmon resonance (SPR). In these methods, a library of antibody or protein binder variants is generated in which individual amino acid residues within one or more complementarity-determining regions (CDRs) or paratope, framework regions, Fc domains, or other binding-relevant domains are systematically substituted, mutated, or modified, while the remaining antibody sequence is held constant. Each variant is immobilized or expressed on a solid support and interrogated for binding to a target molecule, and / or screened sequentially against deep-mutational-scanned variants of target molecule (as analyte flow over on-chip antibody variant library), under near identical assay conditions, yielding kinetic parameters including association rate (ka), dissociation rate (kd), and equilibrium dissociation constant (KD) for each residue pairs (on target and antibody or binder). The change in Gibbs free energy of binding (AG) for each variant is calculated from the measured KD values, and the residue-specific energetic contribution (AG_residue) is determined by comparing the AG of each single-residue variant to that of a corresponding reference antibody sequence. In certain embodiments, the resulting AG-residue values are mapped across the antibody sequence to generate a quantitative energetic landscape of the binding interface, enabling identification of energetically favorable, neutral, or unfavorable residues. These residue-level energetic profiles may be used to guide antibody optimization, affinity maturation, de-liabilization, developability, epitope refinement, or computational and machine-learning-based structure-function modeling of antibody-target interactions.

[0278] In certain embodiments, experimentally derived binding free energy values (AG) and changes in binding free energy (AAG) are used as training data, labels, or target variables for one or more machine learning (ML) and generative artificial intelligence (Al) models configured to analy ze or predict antibody target interactions. The AG values may be calculated from equilibrium binding affinities (KD) measured for antibody variants using kinetic assays, while AAG values are determined by comparing the AG of a variant antibody sequence to that of a corresponding reference sequence differing by one or more amino acid substitutions. In some embodiments, AAG values are computed for single-residue variants, thereby generating residue-specific energetic contributions that serve as supervised learning labels associated with individual amino acid positions, sequence features, or structural descriptors.

[0279] In further embodiments, the machine learning or Al models are trained using input features comprising antibody sequence information, residue identity, positional context, physicochemical properties, structural annotations, or combinations thereof, together with corresponding AG or AAG values as quantitative outputs. The trained models may leam relationships between sequence-level or residue-level features and experimentally measuredAtty. Dkt. No. 155529.00033

[0280] energetic effects on binding, enabling prediction of binding affinity changes, identification of energetically favorable substitutions, or ranking of candidate antibody variants. In certain implementations, the models are further trained across multiple targets, assay conditions, or experimental replicates, allowing the models to generalize energetic patterns and to account for context-dependent or non-additive effects. The resulting Al models may be used to design, optimize, or prioritize antibody sequences with improved binding properties, reduced liabilities, or desired energetic profiles prior to experimental validation.

[0281] Polyreactivity

[0282] In certain embodiments, systems and methods are provided for assessing polyreactivity of drug candidates by screening an on-chip library of drug candidates with one or more polyreactivity agents. In such systems and methods, a plurality of drug candidates are immobilized on a solid support, and polyreactivity agents selected to represent chemically or biologically diverse, non-specific binding partners are introduced as analytes under controlled assay conditions. Binding responses between each immobilized drug candidate and the polyreactivity agents are measured to identify non-specific or promiscuous interactions. Drug candidates exhibiting elevated or reproducible binding signals to one or more polyreactivity agents are classified as polyreactive and may be deselected or removed from further consideration, whereas candidates exhibiting minimal or no binding are classified as having low polyreactivity. The screening may be performed in single or multiple assay cycles, optionally with regeneration steps between cycles, enabling comparative and quantitative assessment of polyreactivity across the on-chip library. This approach supports early identification and de-risking of drug candidates with undesirable non-specific binding properties and facilitates prioritization of candidates with improved developability profiles.

[0283] “Polyreactivity reagents” can be any agent selected to represent non-specific binding partners to the drug candidate. Exemplary polyreactivity reagents include, but are not limited to, albumins, liposaccharides, DNA, and glycan reagents.

[0284] Screening on-chip library against targets variants or isoforms

[0285] In certain embodiments, systems and methods are provided for screening an on-chip library of drug candidates against a target molecule and a plurality of related molecular variants to characterize binding specificity or breadth. Such variants may include naturally occurring or engineered isoforms of the target, disease-relevant mutations, post-translationally modified forms, truncated forms, or homologous proteins belonging to the same target family. In such systems and methods, the target and its variants are introduced sequentially with or without regeneration, or in parallel, as analytes to interact with the immobilized drug library, andAtty. Dkt. No. 155529.00033

[0286] binding responses are measured under substantially identical assay conditions. Comparative analysis of the binding profiles enables identification of drug candidates that selectively bind a desired wild-type or variant target with minimal cross-reactivity, as well as identification of drug candidates that bind broadly across multiple target variants or family members, according to the desired therapeutic effect. This screening strategy supports rational selection and optimization of drug candidates based on the intended therapeutic or diagnostic modality, including highly specific binders or broadly reactive binders, and enables early assessment of selectivity, cross-reactivity, and robustness to target heterogeneity.

[0287] In certain embodiments, methods are provided for identifying drug candidates that preferentially bind to post-translationally modified (PTM) forms of a target relative to the corresponding non-modified target. Such methods comprise screening an on-chip library of binders against a PTM-modified target and a non-PTM target under substantially identical assay conditions and comparing binding responses. Drug candidates exhibiting increased affinity, reduced dissociation, or selective binding to the PTM-modified target are identified as PTM-selective binders. This approach enables discrimination between modificationspecific and modification-independent binding and supports development of reagents or therapeutics targeting PTM-defined disease states.

[0288] In certain embodiments, methods are provided for identifying drug candidates that preferentially bind to target molecules carrying mutations at one or more specific residues relative to the corresponding wild-type target. Such methods comprise screening an on-chip library of binders against mutant and wild-type forms of the target under substantially identical assay conditions and comparing binding responses, kinetic parameters, or competition profiles. Drug candidates exhibiting increased affinity, reduced dissociation, or selective binding to the mutant target are identified as mutation-selective binders, while candidates exhibiting comparable binding to both forms are classified as mutation-tol erant or broadly reactive. This approach enables systematic identification and prioritization of binders that selectively recognize disease-associated or resistance-associated target mutations and supports development of precision therapeutics and diagnostics targeting genetically defined disease states.

[0289] Screening on-chip library against target homologs in animal species

[0290] In certain embodiments, systems and methods are provided for screening an on-chip library of drug candidates against homologs of a target molecule derived from one or more animal species used in preclinical studies. Such homologs may include orthologous proteins from rodent, non-human primate, or other relevant animal models. In embodiments, the animalAtty. Dkt. No. 155529.00033

[0291] model can be mice, rats, dogs, rabbits, non-human primates (monkeys), or organisms like zebrafish, fruit flies, and nematodes. In such systems and methods, the animal-derived target homologs are introduced sequentially with or without regeneration, or in parallel, as analytes to interact with the immobilized drug candidate library, and binding responses are measured under substantially identical assay conditions. Comparative analysis of binding profiles across species enables identification of drug candidates that exhibit conserved binding to both the human target and one or more animal homologs, thereby informing selection of appropriate animal models for pharmacology, efficacy, immunogenicity or toxicology studies. Alternatively, the screening may identify species-specific binding differences that guide model selection, protein engineering, or study design. This approach supports early de-risking of preclinical development by aligning drug candidates with suitable animal models based on target engagement.

[0292] Screening on-chip library for off-target binding

[0293] In certain embodiments, systems and methods are provided for screening an on-chip library of drug candidates to assess off-target interactions. Such screening may comprise exposing the immobilized drug library to one or more panels of proteins representing specific off-target classes or target families, including but not limited to kinases, phosphatases, proteases, GPCRs, ion channels, transporters, transcription factors, receptor families or other functionally or structurally related protein groups. Binding responses are measured for each drug candidate under substantially identical assay conditions to identify unintended interactions with proteins outside the intended target or target class. Drug candidates exhibiting minimal off-target binding are prioritized, while candidates displaying significant crossreactivity are flagged for de-prioritization or further optimization.

[0294] In further embodiments, off-target screening is performed using complex biological mixtures, including but not limited to cell lysates, tissue lysates, serum, plasma, or fractions thereof, or using large-scale protein collections representing a substantial portion of a proteome. In such methods, the complex mixture is introduced as an analyte to the immobilized drug library, and non-specific or promiscuous binding interactions are detected through aggregate binding signals, competition effects, or changes in baseline response. Comparative analysis across drug candidates enables identification of molecules with elevated non-specific binding propensity or undesirable interactions with endogenous biomolecules. This approach supports early assessment and mitigation of off-target liabilities, improves developability and safety profiles, and enables selection of drug candidates with enhanced specificity and reduced risk of adverse effects.Atty. Dkt. No. 155529.00033

[0295] Stability attributes

[0296] In certain embodiments, antibody-based and other biologic drug candidates may exhibit liabilities that adversely affect their development, manufacturing, or clinical performance. Such liabilities may include suboptimal binding affinity or specificity, crossreactivity with unintended targets, or binding to conserved or off-target epitopes that result in undesirable biological effects. Additional liabilities may arise from limited stability under physiological or stress conditions, including susceptibility to thermal denaturation, pH-induced conformational changes, oxidation, deamidation, or aggregation. These factors may lead to reduced potency, altered pharmacokinetics, or loss of functional activity over time. In exemplary embodiments, the on-chip drug library may be screened with one or more reagents or analytes to identify one or more liabilities to deselect the drug variant from further consideration.

[0297] 1. Temperature

[0298] In certain embodiments, systems and methods are provided for assessing thermal stability of drug candidates immobilized in an on-chip drug library. Such system and methods comprise subjecting the immobilized drug candidates to elevated temperature conditions, including incubation at temperatures of about 60 °C to about 90 °C, or approximately 75 °C, or at 150 °C, for a defined period of time, followed by cooling and subsequent screening with the target molecule under assay-compatible conditions. Binding responses obtained after thermal exposure are compared to baseline or pre-incubation binding responses to evaluate loss or retention of binding activity. A reduction in binding signal, altered kinetics, or increased dissociation is indicative of thermal degradation, misfolding, or denaturation of the drug candidate, whereas retention of binding activity indicates thermal robustness. This approach enables parallel, high-throughput assessment of thermal stability across an on-chip drug library and supports selection of drug candidates with improved stability and developability profiles.

[0299] 2. pH

[0300] In certain embodiments, systems and methods are provided for assessing pH robustness of drug candidates immobilized in an on-chip drug library. Such methods comprise exposing the immobilized drug candidates to buffer solutions spanning acidic and basic pH ranges, including pH values below physiological pH and above physiological pH, for defined incubation periods, followed by returning the library to assay-compatible conditions and screening with a target molecule. In other embodiments, the on-chip library may be screened with target in pH buffer solutions having varied acidic or basic pH values. Binding responses obtained after exposure to, or in the pH-challenged conditions are compared to baselineAtty. Dkt. No. 155529.00033

[0301] binding responses to evaluate loss or retention of binding activity'. A reduction in binding signal, altered kinetics, or increased dissociation is indicative of pH-induced degradation, misfolding, or conformational instability of the drug candidate, whereas preservation of binding activity indicates pH robustness. This approach enables parallel evaluation of pH stability' across an on-chip drug library and supports prioritization of drug candidates with enhanced stability and suitability’ for downstream development.

[0302] 3. Aggregation propensity

[0303] In certain embodiments, aggregation propensity is recognized as a significant liability in the development of antibody-based and other biologic drug candidates. Aggregation may arise from partial unfolding, exposure of hydrophobic regions, chemical degradation, or stress conditions encountered during expression, purification, formulation, storage, or administration. Aggregated species can reduce effective drug concentration, impair target binding, and alter pharmacokinetic behavior, and may also increase the risk of immunogenicity or adverse immune responses. Aggregation propensity' may vary’ among drug candidates with otherwise similar binding properties and is influenced by sequence composition, structural stability, post-translational modifications, and environmental factors such as temperature, pH, and ionic strength. Early identification and minimization of aggregation-related liabilities support selection of biologic drug candidates with improved stability7, safety7, and manufacturability.

[0304] In some embodiments, the biomolecules on the surface comprise antibodies, antibodylike molecules, protein binders, peptides, and cyclic peptide biologies containing disulfide bonds which may or may not be prone to aggregation or misfolding. In some aspects, aggregation propensity7is reduced through varying expression conditions during in vitro transcription and translation. Example conditions that can be altered to reduce aggregation include, but are not limited to, altering the ratio of heavy chain and light chain DNA concentrations within the same nanowell (examples include 1:1 or 1:3 molar ratio of HC: LC) for dual chain constructs such as Fabs and mAbs, altering DNA concentration of single chain constructs such as scFv, VHH (example concentrations include 0.1 nM, 1 nM, 3 nM, and 10 nM), altering expression temperature (example temperatures include 37 °C. 30 °C, 26 °C, 15 °C, 10 °C), altering expression time (example incubation times include 2 h, 3 h, 6 h, 12 h, and 20 h). The result of the expression condition alterations successfully reducing aggregation or misfolding propensity is an increased ratio of soluble to insoluble protein. The increased ratio of soluble protein directly correlates to increased functional activity7as determined by SPRAtty. Dkt. No. 155529.00033

[0305] binding and kinetics data to the target protein for which the binder was designed.(Figs. 44A-44B).

[0306] In some aspects, the on-chip library is overexpressed and an aggregation propensity against a target is determined.

[0307] In some aspects, aggregation propensity can be assessed by co-printing different DNA constructs in the same nanowell, wherein each DNA construct encodes for either the HC or LC with or without an additional detection tag. In some aspects, the HC is printed as a single construct with one tag or as two constructs with two unique tags. In other aspects, the LC is printed as untagged or tagged, or a combination of both approaches. Example tags may include Halotag, His, Flag, Spytag, etc. In some embodiments, one of the tags expressed as fusion proteins is designed to bind to a substrate surface, preferably using covalent capture, while the other tag (protein fusions) remains in solution. In some embodiments, only one tag may interact with the substrate surface. For example, HC-Halotag (capture tag) covalently binds to the substrate surface coated with chloroalkane, while the co-printed HC-Histag (detection tag) in the same nanowell cannot interact with the substrate surface and therefore is only detectable upon misfolding or aggregation with or onto the proteins captured on the substrate surface. Figs. 44A-44B). In some embodiments, aggregation is assessed via SPR by flowing antidetection tag antibodies across the substrate surface and measuring binding levels (RU, Rmax). In some aspects, the absence of signal from the non-covalent detection tag (such as His tag) indicates no aggregation and proper folding, wherein the detection of the non-covalent detection tag (such as His tag) indicates aggregation and / or misfolding, wherein the intensity of this signal is proportional to the amount of aggregation. In some aspects, the result of the SPR aggregation assessment informs on ideal expression conditions to produce functional, properly folded biomolecules with minimal aggregation in in vitro transcription and translation reactions in nanowells for simultaneous capture onto a substrate as described herein.

[0308] In some aspects, aggregation and misfolding can be assessed through gel-based analysis comparing soluble and insoluble fractions, wherein the insoluble fraction is pelleted by highspeed centrifugation post-expression, followed by resuspension in a detergent-containing buffer, wherein the post-centrifugation supernatant is considered the soluble fraction. In some embodiments, the expression and aggregation of proteins in tubes or microwell plates is expected to be lower than that expressed and captured in nanowells. In some embodiments, these fractions are analyzed via gel electrophoresis to determine soluble vs insoluble ratios in each expression condition. In some aspects, the result of this assay informs on the ideal expression conditions to produce functional, properly folded biomolecules with minimalAtty. Dkt. No. 155529.00033

[0309] aggregation in in vitro transcription and translation reactions in nanowells for simultaneous capture onto a substrate as described herein.

[0310] In some aspects, misfolding is corrected through refolding procedures or addition of cofactors to assist in proper folding (i.e. chaperones such as DsbC, DnaK, GroE, PDI; reducing reagents such as GSSG / GSSH, detergents, stabilizers, etc.). In some embodiments, cofactors are added during expression or performed as a treatment of the chip with expressed constructs bound covalently to the surface. In other embodiments, refolding procedures can be performed post-expression or included as additives in the in vitro transcription translation expression mixture during expression. In some aspects, refolding procedures may include exposure of proteins to detergents which disrupt aggregation complexes followed by a second buffer which removes the detergents from the protein. In some embodiments, the disruptor or stabilizer may comprise glycerol, polyethylene glycol, sugars, amino acids (such as Arginine), chaperones (DsbC, DnaK, GroE, PDI, etc.), reducing reagents (GSSG / GSSH, DTT, etc.), detergents (sodium dodecyl sulfate, ethyl trimethyl ammonium bromide, etc.). In some embodiments, detergents can be stripped with buffers containing cyclodextrin.

[0311] In some aspects, different biomolecule formats share a common binding target but require unique expression conditions due to differences in aggregation propensity. In some embodiments, a Fab, scFv, and VHH may or may not require unique conditions for each format to yield functional, soluble (i.e. non-aggregated, properly folded) protein products capable of binding to the common target. In some aspects, corrective measures as described herein to reduce aggregation propensity resulting in a more functional and soluble product are required to enable SPR binding kinetic analysis of the biomolecules to their intended target. For example, Adalimumab (trade name Humira) mAh, Fab, and scFv require unique conditions to produce functional proteins binding to TNFa, while Trastuzumab (trade name Herceptin) Fab, mAb, and scFv can produce functional protein products under the same expression conditions. In some aspects, multiple expression conditions may or may not produce functional protein products capable of binding to the intended target. In other aspects, multiple expression conditions producing functional protein products may result in differential kinetic measurements for the same biomolecule sequence binding to the same intended target, wherein analysis of results from kinetic measurements may or may not inform the ideal expression conditions for optimal protein product functionality' with matched kinetic parameters to prior known values.

[0312] Assessing epitope specific immunogenicity of a lead drug candidateAtty. Dkt. No. 155529.00033

[0313] In further embodiments, biologic drug candidates may present developability and safety-related liabilities, including immunogenicity risks such as formation of anti-drug antibodies, rapid clearance, or immune-mediated neutralization. Identification and mitigation of such liabilities during early-stage screening and optimization are critical to improving the likelihood of successful translation of biologic drug candidates into safe, effective, and durable therapeutic products.

[0314] In certain embodiments, the present disclosure recognizes the potential for anti-drug antibodies (ADA) generated by the immune system in response to administration of antibodybased or other biologic drug molecules. Such ADA responses may arise due to sequence, structural, conformational, aggregation-related, or post-translational features of the biologic drug and can result in binding, neutralization, altered pharmacokinetics, reduced efficacy, or adverse immune reactions. ADA formation may be influenced by factors including epitope composition, exposure of immunogenic regions, stability under physiological or stress conditions, and interactions with endogenous proteins. Identification and mitigation of ADA-related liabilities are therefore important for selection and optimization of biologic drug candidates, and methods that enable assessment of immunogenic potential support development of therapeutics with improved safety, durability, and clinical performance.

[0315] In embodiments, a drug candidate or its domains may be deep mutationally scanned and variants produced on chip and screened with serum, plasma or blood samples collected from animal models in preclinical trials, or from patients in human clinical trials, to identify potential residues / epitopes or structural features on antibody / biologic that elicit immune responses. This allows characterizing immunogenic sequence or structural liabilities of new drug molecules, potentially deselecting them from development pipelines or editing-out or changing the immunogenic sequences / structures to lower immunogenicity and improve potential for clinical trial success.

[0316] Al integrated design and affinity maturation using on-chip libraries Machine learning and generative Al are revolutionizing drug development, with several Al -designed drug molecules now advancing to clinical trials, signaling a transformative era in drug design and discovery. Current ML methods for drug discovery leverage iterative design-build-test-leam (DBTL) cycles to create high-affinity binders for drug targets. Al models can produce millions of initial candidate binders in a single ‘design-phase’ run, that are reduced to a few thousand by in-silico binding predictions. How ever, the subsequent build (synthesis) and test phases remain major bottlenecks due to their ex-silico nature, which is both capital and time-intensive. This capability-gap limits the scalability of DBTL cycles, as only a fraction ofAtty. Dkt. No. 155529.00033

[0317] the candidates can be synthesized and tested experimentally. While powerful, Al models for protein design and binding analysis rely heavily on high-quality, large-scale wet lab data sets to train and refine their predictions, and are thus only as good as the real-world data they are trained on. New wet lab technologies are needed not only to generate data at scale for initial training but also for testing and improvement of antibody and protein binders through iterative cycles.

[0318] For conventional and Al-driven drug discovery workflows, experimental testing methods typically balance between ultra-high-throughput assays, which offer binary yes / no binding data, and more comprehensive techniques that yield detailed kinetic and biophysical data but can only accommodate a limited set of binders. Due to the cost-prohibitive nature of synthesizing and testing thousands of candidates, biophysical characterization is often limited to a few dozen lead molecules, representing a significant forced-reduction in sequence diversity. For example, interrogating binding characteristics using techniques such as surface plasmon resonance (SPR) in a high throughput manner is limited both by the limitations in the build (synthesis) phase as well as requirement for significant quantities of each drug candidate for sensor deposition and testing, limiting throughput. Moreover, ultra-high-throughput screening methods, such as phage and yeast display, often produce lead candidates with redundant sequences, reducing the diversity' of final selections. The application of nextgeneration sequencing (NGS) has demonstrated its ability to identify rare clones missed by traditional pull-down assays, highlighting the inadequacy of current methods in capturing sequence diversity comprehensively. These limitations underscore the need for novel high-throughput wet lab assays capable of producing diverse, high-quality data to better inform candidate selection and advance the integration of Al into drug development pipelines.

[0319] In certain embodiments, affinity maturation of antibody-based or other biologic binding molecules is performed to improve binding interactions with a target molecule. Such affinity maturation may comprise generating libraries of variants derived from a parent molecule through sequence diversification, including amino acid substitutions, insertions, deletions, or combinations thereof in one or more binding regions. The variant libraries may be produced using molecular biology techniques, chemical synthesis, or cell-free or cell-based expression systems, such as phage display or yeast display or mammalian display or ribosome or RNA display libraries. The resulting variants are screened or selected to identify molecules exhibiting improved binding affinity', altered kinetics, enhanced residence time, or improved specificity relative to the parent molecule.Atty. Dkt. No. 155529.00033

[0320] In further embodiments, affinity maturation is guided by quantitative binding measurements and iterative screening cycles. Binding parameters such as equilibrium dissociation constants, association and dissociation rate constants, or competition profiles may be used to rank variants and inform subsequent rounds of diversification and selection. Affinity maturation may be performed in a high-throughput or multiplexed format, including on-chip or surface-based screening, enabling simultaneous evaluation of large numbers of variants under substantially identical assay conditions. This iterative process supports systematic optimization of biologic binders for enhanced potency, selectivity, and developability.

[0321] PTM modifications to Antibody or biologic drug molecules for on-chip libraries In certain embodiments, antibody-based and other biologic drug molecules may undergo post-translational modifications (PTMs) that influence their structure, function, stability, and biological activity. Such PTMs may include, but are not limited to, disulfide bond formation that is critical to proper folding and assembly of antibodies or antibody fragments or antibody like binders, PTMS like glycosy lation, phosphory lation, acety lation, methylation, oxidation, deamidation, amidation, disulfide bond formation or reshuffling, and proteolytic processing. PTMs can affect binding affinity, effector function, serum half-life, immunogenicity, and manufacturability, and may introduce molecular heterogeneity among drug molecules. The nature and extent of PTMs may depend on expression system, culture conditions, and downstream processing. Identification, characterization, and control of PTMs are therefore important considerations in the development and optimization of antibody and biologic drug candidates to ensure consistent performance, safety, and efficacy.

[0322] Disclosed aspects provide methods for integrated on-chip drug discovery of biomolecules or drug molecules, such as peptide- or protein-based drug molecules or their conjugates. In some aspects, the drug molecules are antibodies or antigen-binding proteins.

[0323] One aspect is a method for integrated on-chip drug discovery of peptide- or protein-based drug molecules or their conjugates, the method comprising: (a) synthesizing a plurality of drug molecules and attaching the plurality of drug molecules to a solid support surface at discrete locations on the solid support surface, wherein the plurality of drug molecules are synthesized from nucleic acid sequences in cell-free systems or cell-based systems in sealed wells, wherein the plurality7of drug molecules bind covalently to the solid support surface and form a plurality of on-chip drug molecules; (b) contacting the plurality of on-chip drug molecules on the solid support with one or more target molecules under controlled assay conditions; (c) measuring binding interactions between the drug moleculesAtty. Dkt. No. 155529.00033

[0324] and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or mass spectrometry data, or combinations thereof; and (d) analyzing the binding data to perform one or more drug discovery operations selected from affinity ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping, wherein the generating of the drug molecules and the measuring of the binding interactions are performed on the same solid support surface in an integrated workflow.

[0325] As used herein, a “discrete location” may mean that each location has multiple copies of a single peptide or protein at a single location and multiple copies of a distinct peptide or protein at a different location on the solid surface. There can be thousands of discrete locations on a single solid support. The solid support may be a biosensor surface by linking the solid support surface to a biosensor.

[0326] In aspects of the disclosed methods, the method may comprise performing affinity ranking of a plurality of on-chip drug molecules. Affinity ranking of the drug molecules comprising determining one or more binding parameters selected from the group consisting of an equilibrium dissociation constant, an association rate, a dissociation rate, binding free energy, change in binding free energy', residence time, expression yield, or combinations thereof. In aspects, the methods further comprise affinity ranking the drug molecules based on the one or more binding parameters.

[0327] In aspects of the disclosed methods, the method further comprises assessing aggregation propensity' of the on-chip drug molecules, the assessing aggregation propensity comprising detecting changes in binding signal, binding kinetics, or surface behavior of the on-chip drug molecules. In some aspects, the method further comprises reducing the aggregation propensity of the on-chip drug molecules.

[0328] In some aspects of the disclosed methods, the synthesizing step is performed in wells prior to attaching the drug molecule to the solid support surface and wherein the nucleic acid sequences comprise 1) dual chain constructs comprising a heavy’ chain construct and a light chain construct and / or 2) single chain constructs at DNA concentration of about 0.1 nM to about 10 nM.

[0329] In some aspects of the disclosed methods, the nucleic acid sequences are added as a dual chain construct at a ratio of about 1: 1 to about 1: 3 molar ratio of heavy chain construct concentration to light chain construct concentration.Atty. Dkt. No. 155529.00033

[0330] In some aspects of the disclosed methods, the solid support surface and the plurality of drug molecules are complexed with a detection tag or a ligand for the detection tag to allow the solid support surface to attach to the drug molecule via interaction of the detection tag and the ligand for the detection tag. In some aspects, the detection tag and ligand for the detection tag belong to a system selected from the group consisting of a Halotag system, SNAP -tag system, CLIP tag, ACP tag, and Spytag-Spy catcher system.

[0331] In some aspects of the disclosed methods, the method further comprises regenerating the plurality of on-chip drug molecules on the solid support surface for multiple, successive contacting, measuring, analyzing and regeneration cycles. In some aspects, regenerating the on-chip drug molecules comprises contacting the on-chip drug molecules with a regeneration reagent to remove the target molecules without affecting the on-chip drug molecules covalently bound to the solid support surface, and repeating the contacting, measuring and analyzing steps of the method using a second target molecule to interact with the on-chip drug molecules. In some aspects, the regeneration reagent comprises a buffer having a pH of about 2.0 to about 3.0.

[0332] In some aspects of the disclosed methods, the regenerating step and the contacting, measuring and analyzing steps may be repeated using the on-chip drug molecules and a different target molecule in each repeat of the method using the on-chip drug molecules. In some aspects, the regenerating the on-chip drug molecules is repeated 30 times or more.

[0333] Another aspect is a method of integrated on-chip drug discovery of peptide- or protein-based drug molecules or their conjugates, the method comprising: a) contacting a plurality of on-chip peptide- or protein-based drug molecules attached covalently to a solid support with one or more target molecules, wherein the on-chip drug molecules are soluble and capable of binding their native ligand; b) measuring binding interactions between the drug molecules and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or mass spec data, or combinations thereof; and c) analyzing the binding data to perform one or more drug discovery’ operations selected from affinity ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping.

[0334] In some aspects, the method further comprises regenerating the on-chip drug molecules by a method further comprising: d) stripping the one or more target molecules from the plurality of drug molecules while leaving the drug molecules covalently bound to the solid support surface; and e) repeating steps a) through c) with a second target molecule.Atty. Dkt. No. 155529.00033

[0335] In some aspects, the regenerating the on-chip drug molecules and repeating the contacting, measuring and analyzing steps may be repeated up to 30 times with the on-chip drug molecules.

[0336] Another aspect is a method of integrated on-chip drug discovery of peptide- or protein-based drug molecules or their conjugates, the method comprising: a) contacting a plurality of on-chip peptide- or protein-based drug molecules attached covalently to a solid support with one or more target molecules; b) measuring binding interactions between the drug molecules and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or mass spec data, or combinations thereof; c) analyzing the binding data to perform one or more drug discovery operations selected from affinity ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping; d) stripping the one or more target molecules from the plurality of drug molecules while leaving the drug molecules covalently bound to the solid support surface; e) repeating steps a) through c) with a second target molecule.

[0337] A further aspect is a method for integrated on-chip discovery and characterization of protein- or peptide-based drug molecules that includes: (a) synthesizing and presenting, at defined, discrete locations on a solid support surface or chip surface, a plurality7of molecular variants that differ in amino acid sequence or structure; (b) interrogating the molecular variants on the solid support with one or more target molecules or analytes to generate interaction signals corresponding to binding events; (c) deriving, from the interaction signals, quantitative interaction parameters comprising binding data, kinetic parameters, equilibrium parameters, mass spectrometry data or a combination of these; and (d) computationally processing the quantitative interaction parameters to produce one or more discovery outputs selected from affinity ranking of the variants, identification of competing or non-competing binders, epitope classification, epitope mapping, or high-resolution epitope mapping. In aspects, a plurality of molecular variants are synthesized from respective DNA templates in cell-free systems or cellbased systems in sealed microwells or nanowells, including nano-liter, micro-liter, or pico-liter volume wells. In some aspects, the drug molecules bind covalently to the solid support surface or chip surface. In aspects, the drug variant library can be regenerated after assaying with each analyte or target, for multiple successive assay and regeneration cycles. In some aspects, steps (a) through (d) are performed in an integrated manner on the solid support to enable parallel, high-throughput characterization of the molecular variants.Atty. Dkt. No. 155529.00033

[0338] In embodiments of the aspects described above, the method comprises performing epitope binning by sequentially or simultaneously interrogating a plurality’ of on-chip antibody or biologic variants for competitive or non-competitive binding to each (few or all) of the plurality antibody of biologic variants (introduced as analyte), thereby assigning the variants to epitope bins based on overlapping or distinct binding behaviors.

[0339] In embodiments of the aspects described above, the method comprises performing epitope classification or indirect epitope mapping. In embodiments, epitope classification or indirect epitope mapping comprises measuring binding of on-chip antibody or biologic variants for competitive or non-competitive binding to target, sequentially or simultaneously assayed against antibody or protein binders with known epitopes, for indirect epitope mapping and binning.

[0340] In embodiments of the aspects described above, the method comprises performing high-resolution epitope mapping. In embodiments, high-resolution epitope mapping comprises interrogating binding interactions between on-chip antibody or biologic variants and a plurality7of target variants that differ by one or more amino acid residues, truncations, substitutions, or modifications, thereby resolving epitope boundaries at single-residue or near single-residue resolution.

[0341] In the embodiments and aspects described above, the affinity, ranking, epitope binning, epitope classification, or high-resolution epitope mapping is performed in a high-throughput manner with same on-chip drug variant library7, with or without regeneration, under substantially identical assay conditions, enabling direct comparison of binding behaviors across a large number of antibody or biologic variants.

[0342] The aspects and embodiments described above may further comprise integrating results from the affinity ranking, epitope binning, epitope classification, or high-resolution epitope mapping to select, prioritize, or optimize one or more antibody or biologic candidates for downstream development.

[0343] In embodiments disclosed herein, the methods further comprise affinity, ranking. In embodiments, affinity ranking comprises ranking on-chip antibody, protein, or peptide variant library based on one or more binding parameters selected from association rate, dissociation rate, KD, residence time, AG, or AAG.

[0344] In embodiments disclosed herein, the methods further comprise epitope binning. In embodiments, epitope binning comprises performing epitope binning by measuring competitive or non-competitive binding interactions by screening the on-chip peptide, protein, antibody variant or variant library with the target followed by a purified variant or variantAtty. Dkt. No. 155529.00033

[0345] library, followed by regeneration and repeat cycles with the target and a different purified variant.

[0346] In embodiments disclosed herein, the methods further comprise epitope classification. In embodiments, epitope classification may comprise performing epitope classification by grouping on-chip antibody, protein, or peptide variants based on shared or distinct binding behaviors observed across one or more target variants, domains, or reference binders.

[0347] In embodiments disclosed herein, the methods further comprise epitope mapping. Epitope mapping may comprise measuring binding interactions between on-chip antibody, protein, or peptide variant library and a plurality of target variants comprising single-amino-acid mutationally scanned variants of the target, or tiled multi-amino-acid mutationally scanned variants of the target, truncations, fragments, domains, or engineered target constructs.

[0348] In the methods described above, the target constructs may differ by single amino acid substitutions, thereby enabling single- amino-acid-resolution epitope mapping.

[0349] In embodiments disclosed herein, the methods further comprise off-target screening. Off-target screening may comprise interrogating, evaluating, or assaying on-chip antibody, protein, or peptide variant library against one or more non-target proteins, protein families, homologs, or proteome-derived components.

[0350] In embodiments disclosed herein, the methods further comprise assessing polyreactivity of biomolecules, such as antibody, protein, or peptide. Assessing polyreactivity may comprise measuring binding interactions of on-chip antibody, protein, or peptide variant library with heterogeneous molecular mixtures comprising serum, plasma, cell lysates, or panels of unrelated proteins, glyco-proteins, lipo-proteins, lipids, glycans, metabolites, PTM modified proteins, cell surface proteins.

[0351] In embodiments disclosed herein, the methods further comprise performing thermal stability analysis. Thermal stability analysis may be performed by subjecting on-chip antibody, protein, or peptide variant library to elevated temperatures prior to or during binding measurements and assessing changes in binding behavior.

[0352] In the embodiments and aspects described above, the methods further comprise aggregation propensity assessment of on-chip antibody, protein, or peptide variants or variant library. In embodiments, assessing aggregation propensity’ of on-chip antibody, protein, or peptide variant library' comprises detecting changes in binding signal, kinetics, or surface behavior, or detecting a differently tagged variant co-expressed in the same nanowells, indicative of aggregation under defined assay conditions.Atty. Dkt. No. 155529.00033

[0353] In the embodiments and aspects described above, the methods further comprise evaluating hydrophobicity or hydrophilicity of on-chip drug molecules. In embodiments, evaluating hydrophobicity or hydrophilicity by measuring non-specific binding or surface interactions of the on-chip drug molecules.

[0354] In the embodiments and aspects described above, the methods further comprise evaluating solubility of on-chip drug molecules. In embodiments, evaluating solubility of the variants by assessing binding performance under conditions associated with reduced solubility or precipitation.

[0355] In the embodiments and aspects described above, the methods further comprise evaluating pH stability of the on-chip drug molecules. In embodiments, evaluating pH stability by measuring binding interactions of the variants under acidic, neutral, and basic buffer conditions.

[0356] In the methods described herein, a binding free energy (AG) is calculated from an experimentally measured equilibrium dissociation constant (KD) obtained using a surfacebased kinetic assay. In embodiments, the AG is proportional to a temperature-scaled logarithmic function of the KD.

[0357] In the methods described herein, a change in binding free energy (AAG) is calculated as a difference between a AG value of an antibody variant and a AG value of a corresponding reference antibody sequence differing by one or more amino acid residues.

[0358] In the methods described herein, the AAG values correspond to single-amino-acid substitutions within one or more complementarity-determining regions (CDRs) of the antibody or a protein binder. In embodiments, the AAG values are used as residue-specific training labels for a machine learning model.

[0359] In the methods described herein, a machine-learning model is trained using input features comprising antibody sequence information, residue position, physicochemical properties, or structural descriptors, and output labels comprising AG values, AAG values, or combinations thereof.

[0360] In the embodiments and aspects described above, the methods further comprise retraining or updating the machine-learning model using additional AG or AAG values generated from newly measured KD data obtained for one or more additional antibody variants.

[0361] In the embodiments and aspects described above, a trained machine-learning model is configured to predict a binding affinity, a change in binding affinity, or a residue-specific energetic contribution for an untested antibody sequence.Atty. Dkt. No. 155529.00033

[0362] In the embodiments and aspects described above, measuring the binding interactions comprises obtaining, for each drug molecule, conjugates thereof, or component, a plurality of binding parameters comprising at least two parameters selected from an association rate, a dissociation rate, an equilibrium dissociation constant (KD), a residence time, a binding free energy (AG), or a change in binding free energy (AAG). In embodiments, the methods may further comprise using the plurality of binding parameters as input features, output labels, or both to train one or more machine-learning models configured to analyze or predict binding properties of drug molecules. In some embodiments, the machine-learning models are trained using combinations of kinetic parameters and thermodynamic parameters to leam relationships between molecular sequence features, structural features, or physicochemical descriptors and experimentally measured binding behavior. In some embodiments, the method may further comprise calculating AG values from measured KD values under controlled assay conditions, and calculating AAG values by comparing AG values of drug or variant molecules to a reference molecule, and using the AG and AAG values as quantitative labels for training the machine-learning models. In some embodiments, training the machine-learning models comprises jointly processing two or more of association rate, dissociation rate, KD, residence time, AG, or AAG to capture non-linear or non-additive relationships between the parameters. In some embodiments, the method further comprises predicting, using the trained machinelearning models, at least one of a binding affinity, a kinetic rate parameter, a residence time, or a residue-specific energetic contribution for an unmeasured drug molecule. In some embodiments, the method further comprises iteratively updating or retraining the machinelearning models using additional binding parameters generated from subsequent on-chip measurements of newly designed or selected antibody, protein, or peptide variants. In embodiments, the plurality of binding parameters is obtained under substantially identical assay conditions across the variants, enabling direct comparison and combined use of the parameters for machine-learning training.

[0363] In the embodiments and aspects described above, the methods further comprise performing affinity maturation. In embodiments, affinitv maturation is performed by generating or selecting a drug molecule predicted to exhibit improved binding to a target molecule. In embodiments, generating the drug molecules or variants comprises introducing amino acid substitutions, insertions, deletions, or combinations thereof within complementarity-determining regions (CDRs), framework regions, or other binding-relevant domains. In some embodiments, the method further comprises measuring binding interactionsAtty. Dkt. No. 155529.00033

[0364] of the drug molecules on the solid support to obtain kinetic or equilibrium parameters comprising at least one of association rate, dissociation rate, equilibrium dissociation constant (KD), residence time, binding free energy (AG), or change in binding free energy (AAG). In some embodiments, the method further comprises training or updating one or more machinelearning models using the kinetic or equilibrium parameters to predict affinity improvements associated with sequence modifications. In some embodiments, the machine-learning models are used to rank, select, or design affinity-matured drug molecules prior to experimental validation. In some embodiments, the method further comprises iteratively repeating steps of variant generation, binding measurement, and machine-learning model training to progressively improve binding affinity. In some aspects of the disclosed method, affinity maturation is performed using residue-specific energetic contributions derived from AG or AAG values to guide selection of sequence modifications.

[0365] In certain embodiments, the methods comprise measuring, for each of the plurality of drug molecules or synthesized variants of a library7, two or more binding outcome parameters including ka, kd, KD. and residence time, and associating the parameters with the corresponding variant sequence. In certain embodiments, the methods comprise generating, for each drug molecule or variant sequence, a predicted structural representation and storing the predicted structural representation together with a confidence metric in a data storage device or system. In certain embodiments, the methods comprise extracting structural descriptors from the predicted structural representation including one or more of solvent exposure, electrostatic surface features, loop conformations, paratope geometry, or contact maps. In certain embodiments, the methods comprise computing correlations between (i) sequence features and / or structural descriptors and (ii) measured binding outcomes, wherein the correlations comprise residue level attributions, motif level attributions, or region level attributions. In certain embodiments, the methods comprise performing at least one perturbation experiment selected from alanine scanning, saturation mutagenesis, or combinatorial scanning, measuring a change in kd and / or KD relative to a reference, and correlating the change with perturbed residues and local structural context. In certain embodiments, the methods comprise measuring binding outcomes for each drug molecule or variant against a plurality of target moleculess including a primary target molecule and at least one homolog, ortholog, or off target molecule, and computing a selectivity7metric based on differential KD and / or differential kd. In certain embodiments, the methods comprise selecting a preclinical species model based on a model score derived from conservation of measured binding outcomes across species homologs. In certain embodiments, the methods comprise performing competitive binding classification orAtty. Dkt. No. 155529.00033

[0366] epitope binning and correlating bin assignments with sequence and / or structural features to identify epitope class-specific determinants of kd and / or KD. In certain embodiments, the methods comprise performing epitope mapping at single amino acid resolution and storing, for each epitope residue, a corresponding effect on kd, KD, or residence time, thereby generating an epitope resolved sequence-structure-function dataset. In certain embodiments, the methods comprise evaluating developability by conducting one or more assays comprising polyreactivity screening, non-specific binding in serum / plasma, thermal challenge, pH challenge, aggregation assessment, or solubility assessment and correlating developability readouts with sequence and structural descriptors. In certain embodiments, the methods comprise training or fine tuning a machine learning model using training samples comprising sequence, structural descriptors, measured binding outcomes. The model may be configured to predict at least one of ka, kd, KD, residence time, or selectivity. In certain embodiments, the methods comprise selecting subsequent variant sequences of the drug molecules or the target molecules to synthesize using active learning, wherein selection is based on a model uncertainty estimate and / or expected improvement in kd, KD, or selectivity. In certain embodiments, the methods comprise generating a multi objective optimization score for each variant using a weighted function of kd, KD, residence time, AG and / or AAG values, selectivity, and developability, and selecting at least one lead drug molecule or variant thereof based on the score. In certain embodiments, the methods comprise receiving candidate sequences for the drug molecules or the target molecules from a traditional discovery source comprising hybridoma output or display library output, and using computed sequence-structure-binding correlations to design an affinity matured sequence set for synthesis and measurement on the chip. In certain embodiments, the methods comprise receiving candidate sequences for the drug molecules or the target molecule from an Al generative model, measuring binding outcomes on the chip, and providing the measured outcomes as feedback to retrain or update the Al generative model. In certain embodiments, the methods comprise generating a residue level attribution map that identifies residues predicted to increase binding affinity of the drug molecule to the target molecule by decreasing kd and / or decreasing KD, and outputting the attribution map to guide subsequent design.

[0367] In certain embodiments, the methods comprise synthesizing a plurality of drug molecules or variants on a solid support and storing a digital representation of each primary amino acid sequence. In certain embodiments, the methods further comprise generating a predicted structural representation for each drug molecule or variant based on the primary amino acid sequence. In certain embodiments, the methods comprise extracting structuralAtty. Dkt. No. 155529.00033

[0368] descriptors including one or more of surface exposure, electrostatics, secondary structure, or paratope geometry. In certain embodiments, the methods comprise measuring functional binding outcomes for each drug molecule or variant using a biosensor. The functional binding outcomes may comprise at least two of ka, kd, KD, or residence time. In certain embodiments, the methods comprise correlating the measured functional binding outcomes with both the primary’ sequence and the predicted structural representation to generate sequence-structure-function relationships. In certain embodiments, the methods comprise identify ing residue-level or motif-level determinants of binding by associating localized sequence or structural variations with changes in kd, KD, or residence time. In certain embodiments, the methods comprise measuring binding outcomes of the drug molecule or variants against a plurality' of related target molecules and computing selectivity metrics based on differential binding kinetics. In certain embodiments, the methods comprise correlating selectivity metrics with sequence and structural features to identify determinants of target molecule specificity’. In certain embodiments, the methods further comprise evaluating one or more developability attributes selected from polyreactivity’, non-specific binding, thermal stability, pH stability, aggregation propensity', or solubility, and associating the attributes with sequence-structure-function relationships. In certain embodiments, the methods comprise computing a multiobjective optimization score for each drug molecule or variant based on binding kinetics, selectivity, and developability attributes. In certain embodiments, the methods comprise training or updating a machine-learning model using training data comprising primary sequence features, structural features, and measured functional binding outcomes. In certain embodiments, the methods comprise predicting binding outcomes for untested sequences (untested drug molecules or target molecules) using the machine-learning model. In certain embodiments, the methods comprise selecting new drug molecules or variant sequences for synthesis based on predicted improvement or uncertainty, thereby forming an active learning design-build-test-leam cycle. In certain embodiments, the candidate drug molecules or variant sequences are derived from a traditional discovery’ workflow selected from immunization, hybridoma generation, or display-based libraries. In certain embodiments, the candidate drug molecules or variant sequences are generated by an Al-based generative model and experimentally validated using the measured binding outcomes.Atty. Dkt. No. 155529.00033

[0369] Miscellaneous

[0370] The disclosed subject mater may be further described using definitions and terminology as follows. The definitions and terminology used herein are for the purpose of describing particular embodiments only and are not intended to be limiting.

[0371] As used in this specification and the claims, the singular forms “a,” “an,"’ and “the” include plural forms unless the context clearly dictates otherwise. For example, the term “a substituent” should be interpreted to mean “one or more substituents,” unless the context clearly dictates otherwise.

[0372] As used herein, “about”, “approximately,” “substantially,” and “significantly” will be understood by persons of ordinary skill in the art and will vary to some extent on the context in which they are used. If there are uses of the term which are not clear to persons of ordinary skill in the art given the context in which it is used, “about” and “approximately” will mean up to plus or minus 10% of the particular term and “substantially” and “significantly” will mean more than plus or minus 10% of the particular term.

[0373] As used herein, the terms “include” and “including” have the same meaning as the terms “comprise” and “comprising.” The terms “comprise” and “compnsing” should be interpreted as being “open” transitional terms that permit the inclusion of additional components further to those components recited in the claims. The terms “consist” and “consisting of’ should be interpreted as being “closed” transitional terms that do not permit the inclusion of additional components other than the components recited in the claims. The term “consisting essentially of’ should be interpreted to be partially closed and allowing the inclusion only of additional components that do not fundamentally alter the nature of the claimed subject mater.

[0374] The phrase “such as” should be interpreted as “for example, including.” Moreover, the use of any and all exemplary language, including but not limited to “such as”, is intended merely to beter illuminate the invention and does not pose a limitation on the scope of the invention unless otherwise claimed.

[0375] Furthermore, in those instances where a convention analogous to “at least one of A, B and C, etc.” is used, in general such a construction is intended in the sense of one having ordinary skill in the art would understand the convention (e.g., “a system having at least one of A, B and C” would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together.). It will be further understood by those within the art that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the description or figures, should be understood to contemplate the possibilities of including one of the terms, either of the terms,Atty. Dkt. No. 155529.00033

[0376] or both terms. For example, the phrase “A or B’' will be understood to include the possibilities of “A” or ‘B or ‘“A and B.”

[0377] All language such as “up to,” “at least,” “greater than,” “less than,” and the like, include the number recited and refer to ranges which can subsequently be broken down into ranges and subranges. A range includes each individual member. Thus, for example, a group having 1-3 members refers to groups having 1. 2, or 3 members. Similarly, a group having 6 members refers to groups having 1, 2, 3, 4, or 6 members, and so forth.

[0378] The modal verb “may” refers to the preferred use or selection of one or more options or choices among the several described embodiments or features contained within the same. Where no options or choices are disclosed regarding a particular embodiment or feature contained in the same, the modal verb “may” refers to an affirmative act regarding how to make or use and aspect of a described embodiment or feature contained in the same, or a definitive decision to use a specific skill regarding a described embodiment or feature contained in the same. In this latter context, the modal verb “may” has the same meaning and connotation as the auxiliary verb “can.”

[0379] In the foregoing description, it will be readily apparent to one skilled in the art that varying substitutions and modifications may be made to the invention disclosed herein without departing from the scope and spirit of the invention. The invention illustratively described herein suitably may be practiced in the absence of any element or elements, limitation or limitations which is not specifically disclosed herein. The terms and expressions which have been employed are used as terms of description and not of limitation, and there is no intention that in the use of such terms and expressions of excluding any equivalents of the features shown and described or portions thereof, but it is recognized that various modifications are possible within the scope of the invention. Thus, it should be understood that although the present invention has been illustrated by specific embodiments and optional features, modification and / or variation of the concepts herein disclosed may be resorted to by those skilled in the art, and that such modifications and variations are considered to be within the scope of this invention.

[0380] Citations to a number of patent and non-patent references may be made herein. The cited references are incorporated by reference herein in their entireties. In the event that there is an inconsistency between a definition of a term in the specification as compared to a definition of the term in a cited reference, the term should be interpreted based on the definition in the specification.Atty. Dkt. No. 155529.00033

[0381] EXAMPLES

[0382] The following Examples are illustrative and should not be interpreted to limit the scope of the claimed subject matter.

[0383] EXAMPLE 1: Production of single-chain antibody variants on a SPR biosensor chip for measuring target binding kinetics and for characterization of antibody paratopes Drug discovery continues to face a staggering 90% failure rate, with many setbacks occurring during late-stage clinical trials. To address this challenge, there is an increasing focus on developing and evaluating new technologies to enhance the "design" and "test" phases of antibody-based drugs (e g., monoclonal antibodies, bispecifics, CAR-T therapies, ADCs) and biologies during early preclinical development, with the goal of identifying lead molecules with a higher likelihood of clinical success. Artificial intelligence (Al) is becoming an indispensable tool in this domain, both for improving molecules identified through traditional approaches and for the de novo design of novel therapeutics. However, critical bottlenecks persist in the "build" and "test" phases of Al-designed antibodies and protein binders, impeding early preclinical evaluation. While Al models can rapidly generate thousands to millions of putative drug designs, technological and cost limitations mean that only a few dozen candidates are typically produced and tested. Drug developers often face a tradeoff between ultra-high- throughput wet lab methods that provide binary yes / no binding data and biophysical methods that offer detailed characterization of a limited number of drug-target pairs. To address these bottlenecks, the development of the Sensor-integrated Proteome On Chip (SPOC®) platform was previously reported, which enables the production and capture-purification of 1,000 - 2,400 folded proteins directly onto a surface plasmon resonance (SPR) biosensor chip for measuring kinetic binding rates with picomolar affinity resolution. In this study, the SPOC technology is extended to the expression of single-chain antibodies (sc-antibodies), specifically scFv and VHH constructs. These constructs are demonstrated to be capture-purified at high levels on SPR biosensors and retain functionality as shown by the binding specificity to their respective target antigens, with affinities comparable to those reported in the literature. SPOC outputs comprehensive kinetic data including quantitative binding (Rmax), on-rate (ka), off- rate (kei), affinity' (Ko), and half-life (t, 12), for each of thousands of on-chip sc-antibodies.Atty. Dkt. No. 155529.00033

[0384] Additionally, a case study is presented showcasing single amino acid mutational scan of the complementarity-determining regions (CDRs) of a HER2 VHH (nanobody) paratope. Using 92 unique mutated variants from four different amino acid substitutions, critical residues are pinpoint within the paratope that could further enhance binding affinity. This study serves as a demonstration of a novel high-throughput approach for biophysical screening of hundreds to thousands of single chain antibody sequences in a single assay, generating high affinity resolution kinetic data to support antibody discovery and Al-enabled pipelines.

[0385] Introduction

[0386] Biopharmaceutical companies invest billions of dollars annually in developing new drugs to prevent and treat diseases, yet only about 10% of drug candidates that enter development pipelines and clinical trials ultimately receive regulatory approval12The majority of failures are attributed to inadequate efficacy and / or safety, which are often linked to the on- and off-target binding properties of drug molecules. As a result, there is a renewed emphasis on designing new molecular entities (NMEs), particularly biologies, that exhibit high affinity, specificity, and selectivity to their intended targets in early preclinical phase, towards furthering the candidates with improved profiles along the development pipelines. Currently, deep characterization of binding kinetics is typically performed during the lead selection process, and safety assessments are often conducted during costly in vivo animal studies (FIGS. 17A-17D). To improve success rates, it is critical to assess kinetic binding and safety profiles earlier in the preclinical workfl ow-ideally in the design phase-while striving to maintain a broader diversity of lead candidates. Achieving this requires new or improved methods to accelerate the design, build and test cycles, for iterative improvement and down selection of most optimal lead candidates. Such advancements could significantly enhance clinical success rates, saving biopharma hundreds of millions of dollars annually, and expediting the delivery of new and improved therapies to patients.

[0387] Machine learning (ML) and generative artificial intelligence (Al) are revolutionizing drug development, with several Al-designed drug molecules now advancing to clinical trials, signaling a transformative era in drug design and discovery. Current ML methods for drug discovery leverage iterative design-build-test-learn (DBTL) cycles to create high- affinity binders for drug targets(FIG. 18). Al models can produce millions of initial candidate binders in a single 'design-phase' run, that are reduced to a few thousand by in-silica binding predictions. However, the subsequent build (synthesis) and test phases remain major bottlenecks due to their ex-silica nature,Atty. Dkt. No. 155529.00033

[0388] which is both capital and time-intensive. This capability-gap limits the scalability of DBTL cycles, as only a fraction of the candidates can be synthesized and tested experimentally. While powerful, Al models for protein design and binding analysis rely heavily on high-quality, large-scale wet lab data sets to train and refine their predictions, and are thus only as good as the real-world data they are trained on. New wet lab technologies are needed not only to generate data at scale for initial training but also for testing and improvement of antibody and protein binders through iterative cycles.

[0389] For conventional and Al-driven drug discovery workflows, experimental testing methods typically balance between ultra-high-throughput assays, which offer binary7yes / no binding data, and more comprehensive techniques that yield detailed kinetic and biophysical data but can only accommodate a limited set of binders. Due to the cost- prohibitive nature of synthesizing and testing thousands of candidates, biophysical characterization is often limited to a few dozen lead molecules, representing a significant forced-reduction in sequence diversity. For example, interrogating binding characteristics using techniques such as surface plasmon resonance (SPR) in a high throughput manner is limited both by the limitations in the build (synthesis) phase as well as requirement for significant quantities of each drug candidate for sensor deposition and testing, limiting throughput. Moreover, ultra-high-throughput screening methods, such as phage and yeast display, often produce lead candidates with redundant sequences, reducing the diversity of final selections. The application of nextgeneration sequencing (NGS) has demonstrated its ability to identify rare clones missed by traditional pull-down assays, highlighting the inadequacy of current methods in capturing sequence diversity comprehensively 3 6 These limitations underscore the need for novel high-throughput wet lab assays capable of producing diverse, high-quality data to better inform candidate selection and advance the integration of Al into drug development pipelines.

[0390] The class of antibody therapeutics and biologies has undergone significant innovation over the past two to three decades, driving the development of newer therapeutic modalities and yielding highly effective drugs, even as we enter the new age of Al-driven drug design. Following the success of full-length antibody therapeutics such as trastuzumab (Herceptin), bevacizumab (Avastin), and adalimumab (Humira) at the turn of the century7, the therapeutic landscape has expanded dramatically to include a diverse and powerful array of novel singlechain antibody derivatives. These include single-chain antibody formats such as single-chain variable fragments (scFvs), single variable domain of heavy-chain antibodies (VHH, alsoAtty. Dkt. No. 155529.00033

[0391] known as nanobodies, which is a trademark of Ablynx N. V), and shark-derived variable new antigen receptors (VNARs); this is in addition to antibody mimetics such as designed ankyrin repeat proteins (DARPins), and a diverse class of Al designed protein binders. Single-chain antibodies, in particular, offer several advantages due to their smaller size, which often enhances tissue penetration and have fewer functional components outside of the antigen binding region compared to most mammalian antibodies, potentially lowering immunogenicity. These compact structures are also easier to produce using a range of expression systems, including yeast and bacterial platforms such as E.coli.

[0392] Nanobodies, which are derived from the antigen-binding region of specialized singledomain antibodies found in camelids (such as llamas, camels, and alpacas), hold immense promise for current and future clinical applications. These compact molecules, ranging from 12 to 15 kDa in size, consist of a single immunoglobulin domain that exhibits high-affinity binding (nM to pM range) even though it lacks a light chain. The small size and unique structure of nanobodies enable them to access cryptic antigens that conventional full-length antibodies cannot reach. Additionally, in certain modalities, their compact nature can facilitate cellular uptake into the cytoplasm, allowing them to target intracellular antigens?. Like the variable regions of IgG molecules, nanobodies possess three complementarity-determining regions (CDRs) that form the antigen-binding site (paratope).

[0393] However, unlike IgGs, nanobodies are more thermostable, less prone to aggregation, and exhibit a shorter half- life-though the latter can be extended through conjugation with specific proteins. The clinical potential of nanobodies was underscored in 2018 with the FDA approval of Caplacizumab, the first nanobody-based drug for humans, used to treat acquired thrombotic thrombocytopenic purpura8 Caplacizumab demonstrated not only therapeutic efficacy but also low immunogenicity, highlighting its safety’ for clinical use. Since then, nanobody- based therapies have rapidly expanded with a CAR-T cell therapy (Ciltacabtagene autoleucel) incorporating a nanobody-based antigen receptor that was approved in the US and EU in 2022, a nano body for treating solid tumors (Envafolimab) approved in China in 2021, and Ozoralizumab. a nano body to treat rheumatoid arthritis approved in Japan in 20229-11. Numerous others nanobody based therapies are in clinical trials for treatment of infectious disease, autoimmune disease, and cancer.

[0394] Given the therapeutic promise of sc-antibodies and protein binders, drug developers are increasingly focusing on engineering these molecules to address previously encountered challenges related to affinity, specificity, selectivity, and tissue distribution observed in earlier generations of antibody drugs. These efforts aim to enable access to previously obscure targetsAtty. Dkt. No. 155529.00033

[0395] or specific epitopes, significantly expanding therapeutic possibilities. The growing popularity of sc-antibody drug discovery’ is evident from the proliferation of commercially available synthetic libraries, including fully na'fve libraries, designed for rapid target screening and high throughput binder identification. Al-driven engineering of sc-antibodies and protein binders offers the potential to further accelerate the development of new therapies by optimizing paratopes for specific clinical targets. However, realizing this potential requires the development of new wet lab techniques capable of high-throughput, deep characterization of these engineered molecules. Such techniques are essential for rapid testing, analysis, and iterative improvement cycles to identify leads wi th very’ high affinity (or affinity optimized for specific therapeutic modalities), high specificity and selectivity, ensuring their readiness for preclinical and clinical testing.

[0396] As part of the design-build-test-leam (DBTL) cycles in drug development, affinity maturation campaigns are frequently undertaken to iteratively enhance the binding properties of identified binders and lead candidates. The initial step in this process involves deep characterization of the sc-antibody paratopes. This is typically achieved through comprehensive degenerate mutagenesis or computationally driven designer mutations introduced into the CDRs, linkers, and / or scaffold regions. Again, current limitations in the "build" and "test" phases significantly constrain this process. The high cost of producing the library of mutants designed by Al limits the number of sequences that can be experimentally tested for lead selection. As a result, mutations are often restricted to specific substitutions, such as alanine scanning. This narrow approach risks overlooking critical paratope residues essential for binding or alternative amino acid substitutions that could together (cumulative) significantly enhance binding affinity, potentially missing opportunities to achieve sub-picomolar binders and to develop "best- in-class" drug candidates. The limitations of current technologies in generating and testing sufficient sequence diversity can result in suboptimal lead candidates, undermining the potential of affinity maturation campaigns. To address these challenges and bridge the gaps in the "build" and "test" phases, leveraging the previously reported Sensor-integrated Proteome On Chip (SPOC®) platform is proposed. 2

[0397] The SPOC technology offers a transformative solution for high-throughput deep kinetic characterization of sc- antibody (e.g., scFv and VHH) variants, enabling the production of SPR biosensor chips with hundreds to thousands of unique sc-antibody drug candidates produced and captured on chip. The platform allows for direct and simultaneous high-resolution kinetic measurements of analyte binding (e.g., antigen targets in solution) across the entire on-chip scFv or VHH library in a single assay. To address the cost and scalability challenges of theAtty. Dkt. No. 155529.00033

[0398] "build" phase, the SPOC platform utilizes nanoliter-scale cell-free protein expression within high-density nanowells. This approach facilitates the production of thousands of properly folded proteins in discretely separated and isolated nanowells, within a 1.5-square-centimeter area, which are then directly capture-purified onto SPR biosensor chips. The SPOC system enables the characterization of binding interactions for up to 1000- 2400 sc-antibody variants against their antigen target or alternative analytes of choice (e.g., potential off-targets).

[0399] The sc-antibodies are covalently captured on the SPR biosensor chip, allowing for enhanced stability and offering the potential for multiple rounds of regeneration and follow-on assays. This feature ensures that a single chip may be reused for collecting replicate data, further improving cost-efficiency and throughput in the characterization and validation processes. Importantly, the SPOC workflow only requires the DNA sequence library encoding the sc-antibodies or protein binders, eliminating the need for expression and purification of proteins. By leveraging plasmid or linear DNA and cell-free expression systems to convert sc-antibody gene / DNA libraries into protein libraries, the SPOC platform dramatically reduces the time and cost associated with obtaining high affinity resolution SPR kinetic data. Compared to traditional recombinant production approaches, this innovation accelerates the testing and validation process, making it feasible to generate the large-scale, high-quality wet lab data required to train Al models for better prediction accuracies. SPOC technology' thus bridges critical gaps in the "build" and "test" phases, enabling rapid, cost-effective, and scalable characterization of antibody libraries to support the state of art drug discovery pipelines.

[0400] When using Al models for sc-antibody engineering, the SPOC platform addresses critical bottlenecks in the "build" and "test" phases by enabling the interrogation of thousands of nanobody candidates in a single assay. This facilitates the down-selection of the most promising candidates based on objective, high-affinity-resolution kinetic data. SPOC supports affinity maturation cycles by facilitating detailed characterization and engineering of antibody paratopes, enabling the design of variants with improved affinity,, specificity7, and selectivity. With its relatively low production cost, SPOC allows for comprehensive mutational scanning of antibody CDRs, substituting each position with all other amino acids (or a select subset) on a single SPR chip. This approach generates a complete, amino acid-level kinetic dataset to inform the "learn" phase and subsequent iterations of design-build-test-leam (DBTL) cycles. By performing this analysis on the same chip, SPOC ensures direct and confident comparisons of measurements between mutants and the wild-type baseline. The resulting data provide a wealth of information about each variant's binding characteristics, including quantitative binding (Rmax), on-rate (ka), off-rate (kci), affinity (Ko), and half-life (t, 12). Kinetic parameterAtty. Dkt. No. 155529.00033

[0401] ranking enables the selection of binders tailored to specific therapeutic modalities or mechanisms of interest, such as those that bind with highest affinities, or that bind quickly (fast on-rate) at high levels, or that bind very tightly (slow off-rate), or a combination of these. In this study, the application of SPOC technology is demonstrated for the production and analysis of sc-antibodies for the first time. As a use case, the CDRs of a well-characterized HER2 nanobody utilizing single amino acid mutational scanning are analyzed to identify key residues critical for optimizing binding affinity and function.

[0402] Methods

[0403] Materials and Reagents

[0404] Halo-PEG(2)-NH2*HC1 (RL-3680) was sourced from Iris Biotech GrnBH through Peptide Solutions, LLC. Rabbit anti-HaloTag (G9281) and TMR-Halo Ligand (G8251) were obtained from Promega. Mouse anti-HaloTag (#28a8) and Halo VHH (#ot) was obtained from ProteinTech.

[0405] ScFv and VHH antigen targets were sourced as follows: recombinant human TNF alpha (TNFa), Biolegend #570102; recombinant human IL-6, Biolegend #570802; recombinant human HER2 (AA 23-653), Aero Biosystems HE2-H5225; recombinant human CEACAM-5 (CEA), R& D Systems 4128-CM; recombinant human EGFR-Fc, R& D Systems 344-ER-050; recombinant human p53, Active Motif 81091.

[0406] Detection antibodies were sourced as follows: Alexa Fluor®647 Rat anti-human IL-6, Biolegend 501123; Mouse anti-human TNF-a Antibody, BioLegend 502901; Human anti-HER2 (Trastuzumab), Selleck Chemicals A2007; Human anti-CEA (Tusamitamab), Selleck Chemicals A2544; Mouse anti-p53, Sigma P6874; Goat anti-rabbit-Cy3 (Jackson ImmunoResearch 111-165-003), Goat anti-mouse-Cy3 (Jackson ImmunoResearch 115-165-062), Goat anti-Human lgG-Cy3, Jackson ImmunoResearch 109-165-098.

[0407] Identification and construction of single-chain antibody sequences for testing For proper evaluation of single chain antibodies on the SPOC platform, sequences which were well characterized and reported in literature or were otherwise publicly available were tested, such as FDA approved drugs. Sequences were identified using the ExPasy ABCD (AntiBodies Chemically Defined) Database (https: / / web.expasy.org / abcd / ), the International Immunogenetics Information System (IMGT, https: / / www.imgt.org / ), and associated PubMed resources.

[0408] Sequences were designed and codon optimized for cell-free expression using a IVTT E. coli-based kit. These constructs maintain a T7 Promoter-5' UTR-Gene-HaloTag-T7 Terminator structure. The antibody sequences were derived from literature and converted toAtty. Dkt. No. 155529.00033

[0409] scFv format by extracting the VH and VL sequences and separating them with a (Gly4Ser)3 linker. A (Gly4Ser)3 linker was also placed between the scFv sequence and a C-terminal HaloTag. All DNA was synthesized by Twist Biosciences or Alta Biotech.

[0410] The HaloTag was chosen as the fusion tag of choice for these studies due to the covalent bond it forms with its chloroalkane ligand (Halo ligand) that is immobilized on the SPR biosensor surface. The HaloTag is derived from a halogenase enzyme; because the HaloTag protein contains an active site capable of catalyzing a covalent bond between the enzyme's active site and a haloalkane when its properly folded, the use of the HaloTag on the C- terminus of the constructs provides additional confidence that the linker-fused protein of interest is expressed in frame and is properly folded. The underlying rationale is that if the protein-Halo construct is covalently linked to a surface functionalized with chloroalkane (Halo ligand), the HaloTag itself must be folded properly to be enzymatically active for covalent bond formation with the chloroalkane on the sensor surface, and thus, the N- terminal fusion protein is more likely to have undergone proper folding as well.

[0411] In-tube expression of scFvs and nanobodies for validation studies

[0412] E.coli constructs were expressed according to the standard protocol for 6 hours at 37° C, and stored overnight at 4° C prior to assay the next day. For these constructs, a final DNA concentration of 10 nM was used and all reactions were performed in a 96-well PCR plate.

[0413] Validation of protein expression and molecular weight via SDS-PAGE

[0414] All HaloTagged products from E. coli expression reactions were visualized in-gel via tetramethylrhodamine (TMR) fluorescent labeling. Briefly, each cell-free expression reaction was incubated with TMR-Halo ligand in a 1:1:5 ratio of undiluted expression reaction: 50 nM TMR-Halo ligand (Promega): PBS for 15 minutes at room temperature in the dark. Labeling with TMR-Halo ligand prior to denaturation for SDS-PAGE was critical to not destroy the Halo ligand active site / allow the covalent modification of the HaloTagged protein with TMR-Halo. Labeled samples were then prepared in both reducing and non-reducing conditions using Bolt LDS Sample Buffer (Invitrogen) and Bolt Sample Reducing Agent for reducing conditions (Invitrogen), incubated for 5 minutes at 95° C, and loaded into a Bolt Bis- Tris 4-12% gels (Invitrogen). Gels were visualized using the fluorescent protein gel function on a Thermo iBright system (Thermo Fisher Scientific).

[0415] Manual testing of single chain antibody binding

[0416] To analyze the abi li ty of single chain antibodies to bind to their intended target and to measure any off-target binding prior to automated SPR analysis, scFvs and VHH were first tested manually in a multiplexed sandwich assay on glass slides. Glass slides with a hydrogelAtty. Dkt. No. 155529.00033

[0417] coating (Schott Nexterion Slide H) were first functionalized with 4 mM Halo ligand and then blocked with Superblock-TBS (Thermo Fisher Scientific). The slides were rinsed with diH2O and dried under a nitrogen stream, then affixed with a 64-well Flexwell incubation chamber (Grace Bio- Labs). 5 pL of HaloTagged scFv or VHH cell-free expression reaction was added to each well, followed by 15 pL PBST (PBS, pH 7.4, 0.2% Tween-20), and incubated for 1 hat room temperature on a rocker. Enough wells were filled to test each antigen against each scFv or VHH, regardless of specificity, to enable measurement of non- specific binding of antigen.

[0418] Wells were then washed 3 times with PBST, and the target antigen was added at a concentration of 50 nM for most, or 1:20,000 for IgG / lgM-depleted serum (Pel-Freez) for detection of human serum albumin. Negative controls were incubated with PBST instead of antigen. Antigen was incubated for 1 hour at room temperature on a rocker, then washed 3 times with PBST. Detection of antigen bound to scFvs or VHHs was done by incubating with a primary antibody against the antigen for 1 hour at room temperature on a rocker using a 1: 500 dilution in 5% milk in PBST, then washed 3 times in PBST. For any primary antibodies that were not fluorescently labeled, a fluorescently labeled secondary antibody was incubated in the well for 30 minutes using a 1:500 dilution in 5% milk / PBST, then washed 3 times with PBST. One set of wells was also incubated with anti-HaloTag antibody and corresponding secondary antibody to measure total HaloTagged protein bound in the well. Prior to imaging, all slides were gently rinsed with diH2O and dried under a nitrogen stream.

[0419] Slides were imaged with an Innoscan 910AL (Innopsys) using sequential scanning of the 635 nm and 532 nm channels and Mapix software. Quantification was performed using a custom-made GAL file built for the 64-well Flexwell layout, in combination with the Mapix quantification feature. Background signal was subtracted from each well, then normalized to anti-HaloTag signal.

[0420] Preparation and printing of nanowell slides

[0421] Silicon nanowell slides (1" x 3") containing 11,280 individual nanowells were prepared as previously described for DNA printingl2. Briefly, slides were treated with oxygen plasma for 2 minutes at 50 W followed by vapor phase coating with APTES ((3-Aminopropyl)-tri ethoxy silane) and curing for 1 hour at 100° C. DNA was diluted to 100 ng / pL in nuclease-free H2O. Printing of DNA into nanowells was performed with assistance from Engineering Arts, LLC using a Rainmaker 3 piezo-based printer (Bio-Dot) following printing of a print mix composed of bis(sulfosuccinimidyl)suberate (BS3, Thermo Fisher Scientific) and bovine serum albumin (BSA) into each individual well. Printed slides were allowed to dry in theAtty. Dkt. No. 155529.00033

[0422] humidified printing chamber for 15 minutes prior to removal, then stored in desiccating conditions at room temperature prior to assay.

[0423] Capture of protein onto SPR biosensors

[0424] To enable capture of thousands of unique proteins onto planar substrates, SPOC Proteomics designed and built an in-house automated system termed Protein Nano Factory (PNF, previously referred to as AutoCap). The PNF system is capable of introducing cell-free expression lysate into each individual nanowell, incubating the slide for proper expression of protein within each nanowell, and transferring the expressed HaloTagged protein to a planar substrate (glass or biosensor) press-sealed against the nanowell slidel2. The result is a planar substrate covered in discrete, circular spots each containing a pure protein uniquely expressed in each well, with no discernable cross binding between spots 12. Xantec HC30M gold slides were used as the planar substrate for SPR biosensor production. Gold biosensor slides were first prepared for protein capture by activating for 10 minutes with a 1: 1: 1 solution of 0.4M N-(3-dimethylaminopropyl)-N-ethylcarbodiimide (EDC), 0.1 M N-hydroxysuccinimide (NHS), and 0.1 M 2-(N-morpholino) ethanesulfonic acid, pH 5.5 (MES), followed by water rinse and drying with a steady nitrogen stream. Slides were then functionalized with 1 mg / ml Halo-PEG(2)-NH2 overnight at room temperature (Iris Biotech; RL-3680). Any free NHS groups were quenched using 0.5 M ethanolamine, pH 8.5, then gently washed with diH2O and dried under nitrogen prior to loading onto the AutoCap instrument. Printed nanowell slides were prepared for the AutoCap instrument by blocking in SuperBlock-TBS for 30 minutes, then washed with diH2O and dried under nitrogen.

[0425] Following loading of the slides into individual chambers on the PNF system, the chambers were vacuumed and followed with injection of 500 uL of cell-free expression lysate. The nanowell and biosensor slides were then automatically pressed together following injection, isolating each individual nanowell with cell-free expression lysate and DNA to enable protein expression and immediate capture onto the biosensor. The chambers were incubated for 6 hours at 37° C prior to slide removal and immediate rinsing with PBST to remove unbound protein and lysate. SPR biosensors were either loaded immediately onto a custom Carterra LSNr SPR biosensing instrument for equilibration and assay, or stored in 50% glycerol at -20° C for later assay.

[0426] Label-free and multiplex detection of in-solution analytes binding to single chain antibody molecules captured on SPR biosensors

[0427] Biosensors were rinsed with PBS followed by water prior to loading onto the Carterra SPR instrument compatible prism cartridge using 10-15 pL of refractive index-matchingAtty. Dkt. No. 155529.00033

[0428] mounting oil (Cargille). The sensor was then loaded into a custom Carterra SPR instrument and equilibrated overnight in fresh, filtered, and degassed SPR buffer (IX PBS, 0.2% BSA, 0.05% Tween-20, pH 7.2). On this custom instrument, the individual protein spots are visible and can be assigned identifiers using regions of interest (ROI) prior to analyte screening. For validation of protein capture via mouse anti-HaloTag antibody binding, an association time of 6 min and dissociation time of 12 minutes was used. For analysis of analytes binding to single chain antibodies, 15-60 minutes association and 2 minutes dissociation was used. 15 minute association and 20 minute dissociation times were used for calculating affinity constants and other kinetic parameters, in all assays performed in this study.

[0429] Analysis of SPR biosensor data

[0430] Data collected from the custom Carterra LSAxr SPR biosensing instrument was analyzed using Kinetics analysis software (Carterra). Data was processed via y-alignment, double referenced using the leading running buffer injection in addition to empty control spots on the sensor, and erroneous spikes were filtered using standard settings (height= 7, width= 9). Processed data was then globally fit with a 1: 1 Langmuir binding model and kinetic parameters and equilibrium dissociation constants were extracted [on-rate (ka), off-rate (kci), affinity (Ko), Rmax, and half-life (t, 12)]. Mutants with <10% of the maximum signal observed for the highest binder after injection of 400 nM HER2 ECD were considered non-binders for purposes of analysis. Visualization of mutants was performed using PyMol version 3.1.3 based on PDB structure 5MY613.

[0431] Results

[0432] Expression validation of scFv and VHH constructs

[0433] All scFvs and VHHs (Table 1) were designed as C-terminal HaloTag fusion proteins for expression using an E.coli- based IVTT expression kit lysate. All constructs were expressed via IVTT in 96-well PCR plate format and analyzed via SDS-PAGE in reducing and nonreducing conditions to confirm expression and molecular weight, and to evaluate cysteine bond formation. With the exception of non-optimized constructs, TMR-Halo ligand-labeled expression products were easily detected for all constructs at the expected molecular weight(FIG. 22). However, all scFv constructs appeared to produce two distinct expression products. Because both products are also visible in reducing conditions, indicating these are not two different disulfide-bonded species, and because there is no known glycosylation machinery present in the lysate kit, it is likely that a second start site is being used for translation initiation. Interestingly, no additional expression products are observed for equivalents produced with human lysate, despite having the same amino acid sequence (data not shown).Atty. Dkt. No. 155529.00033

[0434] Table 1: List of scFv and VHH constructs tested

[0435] Single Chain Antibody References

[0436] Anti-Interleukin-6 (IL-6) (sirukumab) scFv 14, 15)

[0437] Anti-Human Serum Albumin (HSA) scFv 16

[0438] Anti-Tumor Necrosis Factor alpha (TNFa) VHH 17

[0439] Anti-Human Epidermal Growth Factor Receptor 2 (HER2) VHH 13,18

[0440] Anti-p53 scFv 19

[0441] Anti- Carcinoembryonic Antigen (CEA) scFv 20

[0442] Anti-EGFR (panitumumab) scFv 21

[0443]

[0444] Analysis of constructs synthesized with or without optimization

[0445] Constructs for expression of scFvs in E. coli IVT systems were first designed based on a simple T7 Promoter-RBS- Gene-HaloTag-T7 Terminator structure. Sequences were taken directly from literature and converted to scFv format as required, with no optimization performed on linker length, sequence, or other structural considerations. A standard linker (Gly4Ser)3was used for all scFv constructs tested. These scFv constructs yielded protein in low quantities, but still produced positive binding results when assayed against the scFv targets via fluorescent assay. Further optimization of constructs and expression conditions y ielded at least an 8-fold higher expression of each construct and a corresponding higher level of target antigen bound by each scFv.

[0446] Confirmation of antigen target binding

[0447] ScFv and VHH HaloTag fusions were assessed for binding and specificity for their target antigens in a fluorescent sandwich assay. Antigens bound their corresponding single chain antibodies at various levels but each scFv or VHH showed high specificity for its target compared to all other antigens tested. As expected, little to no signal was observed for the scFv or VHH constructs when antigen was omitted as a negative control nor when the incorrect antigen and corresponding detection antibody was used, confirming specificity. As expected, higher expression of the protein of interest following optimization also improved the total binding signal observed of target antigen.

[0448] Expression and capture of single chain antibodies on multiplexed SPOC SPR biosensors

[0449] High density protein biosensors were prepared for sc-antibody evaluation from nanowell slides printed with DNA encoding constructs designed for expression of sc-antibodies. ScFvs and VHHs expressed using an E. coli IVTT kit resulted in high levels (>200 RU) of protein captured on the surface, as measured with anti-HaloTag antibody (FIG. 23). Recombinant protein antigens matching each of the tested scFv or VHH constructs wereAtty. Dkt. No. 155529.00033

[0450] flowed over the SPR sensor surface and binding levels and kinetics were measured. Multiple single chain antibodies, both in scFv and VHH formats (anti-p53 scFv, anti-HSA scFv, anti-EGFR scFv, anti-CEA scFv, TNFa VHH, and HER2 VHH), were found to bind their target antigen at very high levels in a label-free format (FIG. 24A). Sandwich antibody detection with antigen-specific antibodies for the target antigen were subsequently flowed across the sensor surface and confirmed the presence and specificity of the antigen for its antibody, and further increased the signal of IL-6 binding (FIG. 24C). Specificity of the single chain antibodies and the secondary antibodies for cognate antigens was found to be very high (FIGS. 24B and 24D).

[0451] Affinity' calculations from TNFa and HER2 VHH

[0452] Two constructs (TNFa and HER2 VHH) were chosen for full kinetic analysis. Kinetics were measured on SPOC SPR biosensors in duplicate (spot 1 and 2) by flowing recombinant antigen at multiple concentrations over the sensor surface while collecting kinetic data. Injecting TNFa over the biosensor in increasing concentration from 28 pM to 20 nM using 7 dilutions total resulted in an estimated affinity' (Ko) of< 49 pM. This is about 10-fold higher than the previously reported affinity of 540 pM for this VHH17. Due to the very high affinity of TNFa for the TNFa VHH, this affinity constant is only an initial estimate. This is because limited to no dissociation of bound TNFa was observed during the 30 minute dissociation window. Kinetics measured over a significantly longer dissociation window will be required for more accurate calculation of Ko, and will be performed prior to the submission to a peer review journal. The affinity constant for HER2 binding to the HER2 VHH was calculated to be 11.5 nM using an increasing 5-step titration of HER2 antigen from 1.8 nM to 150 nM (FIG. 25B). This is similar to the reported Koof 4 nM in literature for this VHH.

[0453] Design of single amino acid paratope mutations of HER2 VHH

[0454] Due to the clinical significance of HER2 therapeutic antibodies, the HER2 VHH was chosen for mutational analysis of the CDR regions that make up the paratope (antigen binding region) as a demonstration of SPOC biosensor utility in affinity maturation campaigns. Using data from Mitchell and Colwell22 for manual CDR assignment and automated confirmation via IND123, CDR 1-3 were identified (FIG. 26A) and each amino acid within the CDRs was individually mutated to alanine (substitution to a neutral, non-polar amino acid), aspartate (substitution to a negatively charged amino acid), lysine (substitution to a basic, positively charged amino acid), or serine (substitution to a neutral, polar amino acid), resulting in a total of 92 variants (FIG. 26B). The DNA sequence was kept consistent except for mutations to ensure codon bias did not confound expression. The same codon was used within each amino acid substitution. Each variant was designed with a C-terminal HaloTag for covalent binding to the sensor.Atty. Dkt. No. 155529.00033

[0455] Analysis of HER2 VHH CDR mutagenesis

[0456] SPOC biosensor chips with sc-antibody library comprising mutated CDR variants were prepared from nanowell slides printed with DNA encoding constructs described above, and expressed using E.coli IVTT lysate.

[0457] Recombinant HER2 extracellular domain (ECD) was used as the analyte to measure binding properties of each mutant (FIG. 28A-28B and Supplementary Table 1 of reference 48 in Example 2). Calculation of affinity of wildtype / non-mutated HER2 VHH was consistently measured to be 30 nM ± 4.5 nM from three individual spots. The averaged values from these spots were used for comparison to mutants(FIGS. 29 and 31A-31C). Mutating Y105 to any amino acid resulted in a lower Ko (higher nM), while substitution of D55K resulted in the lowest Ko at 123 nM, ~4 times lower than WT. Altering C33 to any amino acid other than serine ablated binding below the defined detection limit, suggesting this is a critical paratope residue. Alternatively, changing Y28 to any amino acid had no significant effect on the Ko compared to WT, while N 101 D resulted in the highest affinity binder at 21 nM, though only a modest improvement over WT. Analysis of the dissociation rate alone was also performed to most accurately evaluate characteristics of the off- rate (Supplementary Table 2 of reference 48 in Example 2). For variants with an improved off-rate compared to wildtype (n=12), 75% were due to substitution with a charged amino acid, and all but one of these were located in CDR-2 or-3 (Y28K, G54K, G56D, T58K, T58D, N101K, L102K, L102D, T104D). A dissociation rate of 5.7e-5 / s was measured for L102K, similar to a Trastuzumab scFv (kd 5.1e-5 / s, data not shown). Higher resolution sensorgrams are shown for two variants (NIOlKand G54D) with low dissociation and high signal for comparison to wildtype (FIG. 27).

[0458] Analysis of affinity' of HER2 ECD for HER2 VHH produced from a titration of DNA encoding HER2 VHH

[0459] To determine whether evaluated kinetic parameters were dependent on the quantitative amounts of the mutated protein spots across the SOR chip, as each of the protein mutants may express at different levels, an experiment was performed where the printed amount of DNA encoding WT HER2 VHH was titrated into nanowells. Although the magnitude of WT HER2 ECD binding signal increased with printed DNA concentration, the evaluated affinity remained effectively the same. The result confirms that kinetics evaluated for different VHH variants is independent of any variant specific changes in proteins expression levels, over a certain cutoff. This shows that affinity is consistent and independent of levels or variation in protein quantities captured at each spot on the SPR biosensors. Affinity values for HER2 ECD binding to HER2Atty. Dkt. No. 155529.00033

[0460] VHH (wildtype) expressed from DNA printed at 25, 33, 50, 75, and 100 ng / pL were measured to be 30 nM, 30 nM, 35 nM, 34 nM, and 30 nM, respectively, for an average of 31.8 nM ± 2.95 nM.

[0461] Discussion

[0462] A highly multiplexed method for measuring the kinetics of hundreds to thousands of protein interactions on a single biosensor using cell-free expression and high throughput SPR12 previously described. Here, anew application of the SPOC platform (microarray or nanowell array platform or device) is described demonstrating the production and testing of single chain antibodies via in situ cell-free expression and capture-purification onto SPR biosensors, with a use-case application of mutational scanning of an anti-HER2 VHH paratope.

[0463] For the development of this application, fluorescence-based assays were used as an initial test to validate antibody -target binding due to the ability to control a matrix of factors at one time for optimization of dilutions, buffers, and choice of detection antibody with high sensitivity. Following this, a set of sc-antibody molecules were tested using a SPOC chip on a custom Carterra SPR instrument to determine if scFv and VHH antibody fragments produced in cell-free systems are functional and bind to respective targets with high specificity. This facilitates characterizing on-chip sc-antibody selectivity with significantly higher throughput, and to collect kinetic measurements of antigen binding for in-depth characterization and affinity ranking.

[0464] This paper demonstrates that the SPOC Protein Nano Factory (PNF) system produces functional single chain antibodies in the two forms tested directly on SPR biosensors: scFvs derived from full-length antibodies and single-domain antibody VHH. also commonly known as nano bodies. Two versions of a select few of the same scFvs were ultimately expressed using an E. coli-based kit. The first set of E.coli scFvs (non-optimized) produced functional antibody molecules capable of binding target antigen, but expression was low. For the second set of scFvs produced in E. coli, DNA construct and experimental conditions optimization was performed, which resulted in 8-fold more protein on average(FIG. 18). These optimizations produced scFvs that bound highly detectable target antigen on SPOC chips via SPR.

[0465] All subsequent E.coli scFvs and VHH constructs were synthesized post-optimization and their expression products were tested on SPOC biosensors. Expressed protein was confirmed to be present on the sensor at readily detectable levels as measured via anti-HaloTag antibody. It was immediately clear that 4 scFvs (targeting HSA, EGFR, p53, and HSA) and 2 VHH (targeting TNFa and HER2) bound their label-free target antigen at high levels at first assay with no additional optimization of SPR or expression parameters required. These constructs correlated well with previously reported affinity, with these immediately detectable antigens binding to constructs with low to sub-nanomolar affinity.Atty. Dkt. No. 155529.00033

[0466] The constructs tested here had sequences taken directly from literature and were converted to scFvs using a single common linker of (Gly4Ser)3. without experimenting with multiple linker formats or other scFv components or scaffolds. IFNy, CA15.3, TNFa, and other HER2 scFv constructs were also tested using this common linker. CA15.3 and TNFa scFv have shown no binding thus far in this single-shot testing, while IFNg and Trastuzumab- derived HER2 scFv were shown to be capable of binding antigen (data not shown), though selectivity data is pending. Traditionally, scFvs are designed in the format of VH-(Gly4Ser)3-VL, as the most characterized and preferred linker format. However, given the precise scaffolding and folding requirements of scFvs, quite often this requires further optimization with different linker scaffolds if the initial format results in misfolding or aggregation, with no binding of target. Researchers often evaluate different linker lengths, different linker tvpes. switching the order of the VH and VL, moving a purification or expression tag from one terminus to another, etc. Any non-functional scFv constructs with alternative linker designs and other formats will be tested in the future.

[0467] From these immediately functional constructs, tw o VHH w ere further analyzed to calculate affinity and limit of detection via injection of a titration of antigen. For the TNFa VHH, an approximate affinity of <49 pM was calculated, nearly 10-fold higher than the reported affinity of 540 pM for this VHH19. However accurate kinetics could not be measured due to the apparent lack of antigen dissociation over the 20 minute measurement window, which was used for all assays in this study. Further studies need to be performed with a significantly longer dissociation time, to accurately measure dissociation constant and evaluate affinity, and will be done prior to submission of paper to a peer review journal.

[0468] Furthermore, when using SPR biosensing to characterize kinetic parameters of very high affinity binders (pico- molar to femto-molar affinities), the "chaser assay" method may be used to accurately characterize ultra-low-dissociation-rate binding events measured by SPR instrument on SPOC chips. Used in the field of enzymology and more recently described in detail for application to SPR biosensing methodology by Quinn et al [cite], this method uses a "chaser probe" to measure the fraction of free binding sites on the ligand epitope of interest after an acceptable dissociation time (minutes-hours versus many hours-days), resulting in high resolution readout at very high affinity. For the HER2 VHH, the measured affinity was more accurate due to measurable dissociation of the recombinant HER2 extracellular domain over a 20 minute window. This affinity was calculated to be 11.5 nM in this first study, very7close to the previously reported affinity, for this VHH of 4 nM.

[0469] As demonstration for use of the SPOC technology in drug development, a panel of paratope mutants of the HER2 VHH was created and the binding characteristics of HER2 to each mutantAtty. Dkt. No. 155529.00033

[0470] measured simultaneously on a single chip for direct comparison to one another. By mutating each amino acid individually to alanine within the three complementarity determining regions, it can be determined which amino acid side chains are critical for antigen binding. Similarly, by mutating the same residues individually to aspartate (acidic), lysine (basic), or serine (polar), whether changing the side chain properties improves or reduces antigen binding can be determined. Results show that several mutants improved overall affinity', with several improving the off-rate, though some mutants improved affinity by increasing the on-rate. Nearly all improvements in affinity were made by substitution of an uncharged amino acid with a charged amino acid.

[0471] During analysis of the kinetic data from HER2 VHH mutant library', results showed that certain HER2 VHH variants produced binding curves that deviated from the standard 1: 1 binding model used for curve fitting (either due to presence of HER2 dimers or other factors that could have introduced kinetic heterogeneity). To more accurately estimate off-rates, a follow on analysis with the 1:1 binding model to the dissociation phase alone was performed. This analysis resulted in a more accurate fit of the dissociation kinetic curve for better comparison between mutants. As kdis known to be the primary driver of the overall affinity’ of a molecule, especially when stronger binding is desired, an accurate comparison of this constant is required. It was noted that one particular variant, L102K, resulted in a very' low dissociation (kd= 5.7e-5 / s), which was comparable to a trastuzumab anti-HER2 scFv (kd= 5.1e-5 / s) present on the same sensor (data not shown). The variant with the second lowest dissociation rate was notably L102D (kd= l.le-5 / s), indicating this particular residue may be mutated for further improvement of kd, Using data-trained Al-driven models for iterative design on more optimal sequences, 10 -1 OOx improvements in affinity' via modulation of critical amino acids becomes possible.

[0472] There was a notable difference between the first measured affinity of HER2 ECD for HER2 VHH (11.5 nM) and the affinity measured for wildtype replicates in the follow-on second study' that included mutation variants (30 nM). Though the affinities reported are lower than the original report of 4 nM, it well-known that the variation between instruments, especially those produced by different manufacturers, can be as high as 10-fold different24»25. More pertinently, there was one main difference between the first and the second study that may account for the 3-fold difference observed between the two affinities in this study. Of critical difference was the analyte used; though the HER2 ECD used was from the same batch, this recombinant protein was shipped in a buffer containing trehalose as a cry oprotectant which is known to cause a large bulk effect on the SPR signal, thus confounding data. To overcome the possibility of a bulk effect, the protein buffer had to be exchanged into PBS to remove theAtty. Dkt. No. 155529.00033

[0473] trehalose and estimate the final protein concentration using absorbance at 260 nm. This method can be significantly inaccurate, especially for non-lgG proteins, but was used due to a supply shortage at the time of the study. Most critically, the analyte used for the first study was buffer exchanged separately from the second study; since both concentrations were merely an estimate, this is the most likely culprit for the differences in reported affinities between the two studies, as the analyte concentration directly impacts the affinity calculation. The concentration of each buffer exchanged batch will be measured via BCA. Second, the process of buffer exchange may have resulted in differences in protein stability and / or dispersity, with the possibility of some species existing as different multimers if any aggregation occurred as a result of the exchange, or as inactive forms. It is notable, however, that the second study- included 8 replicates of the wildtype HER2 VHH (including the DNA titrationin FIG. 29), and had a low variation in affinity (30 nM ± 3.7 nM, n =8).

[0474] Described for the first time is a method for producing single chain antibodies on SPOC SPR biosensor chips via cell-free protein expression in a highly multiplexed format for label-free detection and characterization of target antigen binding. This method allows the user to simply use a DNA input for biosensor production of thousand(s) of unique single chain antibody constructs, rather than expressing each construct of interest individually at in sufficient amounts followed by spotting or capturing onto a SPR biosensor chip for kinetic assay. SPOC can be applied to evaluate the affinities of antibodies produced via computational modeling to reduce downstream costs and increase the throughput of testing. For example, expression / production of a panel of 92 mutationally scanned variants from the CDR of HER2 VHH on one SPR biosensor is demonstrated, directly from DNA in an E. coli IVTT lysate. The SPOC SPR biosensor, meanwhile, has the capacity to interrogate up to 2,400 mutants, as reported previously. The result is the capability to collect and measure kinetic data at-scale at significantly lower cost per sequence than traditional recombinant production methods. This demonstrates a significant improvement to the 'build' and 'test' cycles required in traditional drug discovery, Al- enabled and Al-driven workflows. Lastly, the results demonstrate how a lead candidate can be mutationally scanned at low-cost and at high-throughput with kinetic measurements to support deep paratope characterization for subsequent affinity maturation campaigns, proposing the application of SPOC platform for iterative DBTL cycles to support traditional and Al drug discovery- pipelines. The future goal for this application will be to fully incorporate SPOC analysis into Al drug discovery workstreams, to demonstrate how the improved "build” and "test’ ’ phases with increased sequence diversity testing capabilities, willAtty. Dkt. No. 155529.00033

[0475] improve the subsequent iterative “learn’" and “design’" phases and result in improved lead drug candidates, ultimately towards improving drug success rates.

[0476] References

[0477] 1. Sun, D., Gao, W., Hu, H. & Zhou, S. Why 90% of clinical drug development fails and how to improve it? ActaPharm Sin B 12, 3049-3062 (2022).

[0478] 2. Smietana, K., Siatkowski, M. & MOller, M. Trends in clinical success rates. Nat Rev Drug Dtscov 15, 379-380 (2016).

[0479] 3. Barreto, K. et al. Next-generation sequencing-guided identification and reconstruction of antibody CDR combinations from phage selection outputs. Nucleic Acids Res 47, e50-e50 (2019).

[0480] 4. Yang, W. et al. Next-generation sequencing enables the discovery of more diverse positive clones from a phage-displayed antibody library. Exp Mol Med 49, e308-e308 (2017).

[0481] 5. Erasmus, M. F. et al. Insights into next generation sequencing guided antibody selection strategies. Sci Rep 13, 18370 (2023).

[0482] 6. Mejias-Gomez. 0. et al. Deep mining of antibody phage-display selections using Oxford Nanopore Technologies and Dual Unique Molecular Identifiers. N Biotechnol 80, 56-68 (2024).

[0483] 7. Sun, X., Zhou, C., Xia, S. & Chen, X. Small molecule-nano body conjugate induced proximity controls intracellular processes and modulates endogenous unligandable targets. Nat Commun 14. 1635 (2023).

[0484] 8. Scully, M. et al. Caplacizumab Treatment for Acquired Thrombotic Thrombocytopenic Purpura. N Engl J Med 380, 335-346 (2019).

[0485] 9. Martin, T. et al. Ciltacabtagene Autoleucel, an Anti-B-cell Maturation Antigen Chimeric Antigen ReceptorT- Cell Therapy, for Relapsed / Refractory Multiple Myeloma: CARTITUDE-1 2-Year Follow-Up. Journal of Clinical Oncology 41, 1265-1274 (2023). 10. Markham, A. Envafolimab: First Approval. Drugs 82, 235-240 (2022).

[0486] 11. Keam, S. J. Ozoralizumab: First Approval. Drugs 83, 87-92 (2023).

[0487] 12. Novel Sensor Integrated Proteome on Chip (SPOC TM) Platform for Evaluating Kinetic Parameters of Protein Interactions in High Throughput.

[0488] 13. D'Huyvetter, M. et al. 1311-labeled Anti-HER2 Camelid sdAb as a Theranostic Tool in Cancer Treatment. Clin Cancer Res 23, 6616-6628 (2017).

[0489] 14. Xu, Z. et al. Pharmacokinetics, pharmacodynamics and safety of a human anti-IL-6 monoclonal antibody (sirukumab) in healthy subjects in a first-in-human study. Br J Clin Pharmacol 72, 270-81 (2011).Atty. Dkt. No. 155529.00033

[0490] 15. Ehrenmann, F., Kaas, Q. & Lefranc, M. IMGT / 2Dstructure-DB card for INN 9431 (sirukumab). IMGT / 3Dstructure-DB https: / / www.imgt.org / 3Dstructure- DB / cgi / details.cgi?pdbcode=9431 (2024).

[0491] 16. Adams, R. et al. Extending the half-life of a fab fragment through generation of a humanized anti-human serum albumin Fv domain: An investigation into the correlation between affinity and serum half-life. MAbs 8, 1336-1346 (2016).

[0492] 17. Beimaert, E. et al. Bivalent Llama Single-Domain Antibody Fragments against Tumor Necrosis Factor Have Picomolar Potencies due to Intramolecular Interactions. Front Immunol 8,867 (2017).

[0493] 18. Vaneycken, I. et al. Preclinical screening of anti-HER2 nanobodies for molecular imaging of breast cancer. The FASEB Journal 25. 2433-2446 (2011).

[0494] 19. Caron de Fromentel, C. et al. Restoration of transcriptional activity of p53 mutants in human tumour cells by intracellular expression of anti-p53 single chain Fv fragments. Oncogene 18, 551-7 (1999).

[0495] 20. Boehm, M. K. et al. Crystal structure of the anti-(carcinoembryonic antigen) singlechain Fv antibody MFE-23 and a model for antigen binding based on intermolecular contacts. Biochem J 346 Pt 2, 519-28 (2000).

[0496] 21. Ehrenmann, F., Kaas, Q. & Lefranc, M. IMGT / 3Dstructure-DB card for 5sx4 (panitumumab). IMGT / 3Dstructure-DB www.imgt.org / 3Dstructure- DB / cgi / details.cgi?pdbcode=5SX4 (2024).

[0497] 22. Mitchell, L. S. & Colwell, L. J. Comparative analysis of nanobody sequence and structure data. Proteins 86, 697-706 (2018).

[0498] 23. Natural Antibody- NANOBODY Database, research.naturalantibody.com / nanobodies (2024).

[0499] 24. Yang, D., Singh, A., Wu, H. & Kroe-Barrett, R. Determination of High-affinity Antibody-antigen Binding Kinetics Using Four Biosensor Platforms. J Vis Exp (2017) doi: 10.3791 Z55659.

[0500] 25. Katsamba, P. S. et al. Kinetic analysis of a high-affinity antibody / antigen interaction performed by multiple Biacore users. Anal Biochem 352, 208-21 (2006).

[0501] EXAMPLE 2: Real time SPR biosensing to detect and characterize fast dissociationrate binding interactions missed by endpoint detection and implications for off-target toxicity screeningAtty. Dkt. No. 155529.00033

[0502] Accurate detection of biomolecular interactions is essential in many areas, from detecting the presence of biomarkers in the clinic, development of therapeutic drugs and biologies in biopharma, to understanding various biological processes in basic research. Traditional endpoint approaches can suffer from false-negative results for biomolecular interactions with fast kinetics. By contrast, real-time detection techniques like surface plasmon resonance (SPR) monitor interactions as they form and disassemble, reducing the risk of falsenegative results. By leveraging cell-free expressed proteins captured on either glass or SPR biosensors and using two different commercial antibodies with variable off-rates that both target HaloTag antigen as a model, results from fluorescence endpoint assay versus real-time sensor-integrated proteome on chip (SPOC®) SPR-based detection is compared and contrasted. In this study, the limitations of the representative immunofluorescent endpoint assay are illustrated when investigating transient interactions characterized by fast dissociation rates. The importance of choosing reagents well-suited to the selected assay are highlighted, as well as the importance of considering binding kinetics and protein ligand conformational states when interpreting results from binding assays, especially for applications as critical as the off-target screening of therapeutics.

[0503] Introduction

[0504] Detection of biomolecular interactions is fundamental to applications in diagnostics, proteomics, and drug discovery. Diagnostics rely on the capture and detection of circulating biomarkers, proteins function within networks of interactions, and therapeutic drugs act by binding to specific targets. To screen for and detect these interactions, traditional methods have leveraged endpoint assays wherein a single measurement is made after a series of incubations and reagent wash steps. However, molecular interactions are not static but are equilibrium reactions. They are rather driven by a dynamic balance between rates of association (ka) and dissociation (kd). Therefore, a critical limitation of end-point assays revolves around the risk of false-negative results when attempting to detect interactions with fast binding kinetics. Such transient interactions may form, yet dissociate rapidly before detection. Endpoint assays are exposed to this limitation because they rely on the bound complex to be stable through a multitude of washing and secondary incubation steps in order for detection to be successful.

[0505] Interrogating transient interactions is of critical importance in drug discovery, where therapeutic specificity is paramount. While generally weaker than the intended on-target binding interactions, transient off-target binding interactions can be significant at elevated drug doses and elevated endogenous expression levels of off-targets in vivo[l]. Small molecule drugs for example have been estimated to interact with a minimum of ~6-l 1 unintended targetsAtty. Dkt. No. 155529.00033

[0506] in the human body [2, 3], Even with therapeutic modalities considered less promiscuous like antibodies, investigations have identified that 33% of lead candidates exhibit off-target binding[4]. This lack of specificity has major implications for therapeutic efficacy and success rates. Approximately 75% of adverse drug reactions (ADRs) are due to dose-limiting toxicity which constrain therapeutic windows [5-7], This toxicity occurs largely due to the interaction of drugs with off-target biomolecules, a problem contributing to an estimated 30% of drug failures.

[0507] In vitro promiscuity correlates with in vivo toxicity. Therefore, secondary pharmacological profiling assays for detecting interaction with a panel of targets most commonly associated with toxicity are required by regulatory guidelines for investigational new drugs and are critical in early-phase drug development] 8- 10|. Pharmaceutical companies employ a battery of secondary pharmacological profiling assays that often consist of panels of putative unsafe off-targets, and rely on various radioligand- or fluorescent-based endpoint detection methods. Examples include the assays offered by the Eurofins Discovery portfolio which are diverse and consist of panels of recombinant proteins, or the Charles River Retrogenix platform, consisting of arrays of cells over-expressing membrane proteins, to name afew[ll-13].

[0508] While useful these endpoint assays provide limited insight into the nature of any given interaction, offering only a narrow snapshot of bound complexes. Furthermore, these endpoint assays risk missing weaker off-target interactions that may nonetheless reduce safety and efficacy when applied in the clinic. Relying solely on such assays may be the difference between identifying a safe and effective therapeutic versus one with un-foreseen off-target binding that fails Phase 1 trials.

[0509] To address these limitations, a variety of real-time and label-free biosensing approaches have been developed including surface plasmon resonance (SPR)[14,15J. In contrast to endpoint assays, SPR can report the interaction with an analyte in solution without the need for fluorescent dyes or other reporter tags (label-free) and importantly, as binding occurs (real-time)-improving the detection of short-lived bound complexes. SPR has become a gold-standard technique for directly measunng the ka and kd of molecular interactions which can be used to calculate occupancy times, bound complex half-life (tl / 2), and the equilibriumdissociation constant (KD)[16,17],

[0510] By leveraging real-time assays in secondary pharmacological screening via techniques like SPR, the risk of false-negatives in off-target binding detection can be reduced; thereby improving the flagging of compounds or biologies with undesired off-target interactions in theAtty. Dkt. No. 155529.00033

[0511] early-phases of drug discovery'. SPR has long-been leveraged in drug discovery pipelines to gain insights into drug-target binding interactions; and now needs to be applied for detection of undesired drug interactions.

[0512] Beyond off-target binding considerations, binding affinity characterization (KD) is also critical in drug discovery. When it comes to affinity, however, more is not always better. This has been the lesson learned in several burgeoning therapeutic modalities such as chimeric antigen receptor T-cell therapy (CAR-T), antibody drug conjugates (ADC), and targeted protein degradation (TPD). Seven CAR-T cell therapies have been approved by the FDA since 2017 and it has been observed that moderate affinity' of the antigen binding domain (KD = -50.0-100 nM range) correlates with antitumor efficacy of CAR-T therapies in the clinic [ 18,19], While the first ADC to gain FDA approval was in the year 2000, next-generation designs have been created to overcome issues with dose limiting toxicity, with nine out of the thirteen FDA-approved ADCs gaining approval recently in just the past five years

[0020] . In ADCs, reducing the affinity for target binding has been found to be a feasible strategy to improve efficacy; yielding increased tumoral diffusion and reduced on-target, off-site related toxicity [21,22], TPD therapies orchestrate ternary complex formation between target proteins and native degradation effector molecules, like E3 ligases of the ubiquitin proteasome system. This ternary' complex triggers choreographed processes that ultimately lead to the elimination of the target protein using natural processes

[0023] , As in CAR-T and ADC therapies, TPD-based therapies require precise affinity tuning to optimize efficacy. Higher affinity of TPD molecules can alter the binding dynamics towards non-functional binary interactions, shifting away from productive ternary complex formation and contributing to the well-recognized “hook effect” [24,25],

[0513] Sensor-integrated Proteome On Chip (SPOC®) is a next-generation protein biosensor technology enabling high density protein production directly onto SPR biosensors for costefficient and high-throughput real-time analyte screening

[0026] , SPOC leverages in vitro transcription and translation (IVTT) on proprietary' Protein NanoFactory' systems to synthesize various proteins of interest fused to a common HaloTag domain, used for in situ capture purification onto chloroalkane coated SPR biosensor slides or glass slide substrates. By coupling cost-efficient cell-free protein synthesis for high density on-chip protein libraries and label-free technologies like SPR, SPOC is poised to improve real-time biomarker screening, kinetic evaluation of therapeutic biologies and drugs, and basic research into protein interaction networks. In this study, two commercial antibodies raised against the common HaloTag protein tag are used to illustrate advantages of real-time screening. Results show that fluorescentAtty. Dkt. No. 155529.00033

[0514] endpoint assay yields disparate binding results between these two antibodies when screening for successful capture of the IVTT protein spots on glass slide substrate. By contrast, real-time screening by SPR demonstrates that both antibodies are similarly capable of binding to the HaloTag fusion proteins present on the biosensor surface and contend that the different kinetic profiles exhibited by these two antibodies result in biased, false-negative binding results when screening by traditional fluorescent endpoint assay. Furthermore. SPOC technology is shown to enhance the multiplex capacity of SPR screening, yielding up to -864 protein ligand spots in a custom LSAXT Carterra instrument (a -2.2-fold increase from the standard 384 commercial instrument capacity).

[0515] Materials and Methods

[0516] Antibodies

[0517] Monoclonal mouse anti-HaloTag (Antibody #1) was sourced in phosphate buffered saline (PBS) only format from Proteintech (28a8) while the polyclonal rabbit anti-HaloTag (Antibody #2) was sourced from Promega (G9281). Goat anti-mouse (115-165-062) and antirabbit (111-165-003) Cy3-labeled secondary’ antibodies were sourced from Jackson ImmunoResearch. In fluorescent assays, all primary antibodies were diluted into IX PBS containing 0.2% Tween-20 and 5.0% fat-free milk (PBST-M).

[0518] In vitro transcription and translation of SPOC arrays

[0519] Capture of the IVTT SPOC arrays was performed as described previously

[0027] , Briefly, plasmid DNA containing HaloTag fusion protein open-reading frames and compatible with cell-free expression were sourced from DNASU plasmid repository and printed into nanowells of a nanowell slide. The nanowell slide was then affixed to SPOC Proteomics’ in house built and proprietary Protein NanoFactory systems (previously AutoCap) along with either glass or biosensor capture slide substrates, depending on the assay. HeLa IVTT cell-free extract (ThermoFisher, 8882) was prepared and injected over the nanowell slide surface, followed by press sealing of the nanowells against the respective capture slide surfaces. This sandwich was incubated at 30 °C for at least 2.0 hours prior to disassembly and rinsing of nanowell and capture slides in IX PBS supplemented with 0.2% Tween-20 (PBST). Capture slide surfaces were subsequently used for follow-on fluorescent or SPR assays.

[0520] Fluorescent assay

[0521] Hydrogel coated partially activated glass capture slides were purchased from Schott (1070936). Amine-terminated HaloTag ligand was purchased from Iris Biotech (RL-3680). Amine-terminated HaloTag ligand was prepared at 1.0 mg / mL concentration and 80 pL of the solution was pipetted onto a clean lifter slip followed by placement of the glass capture slide facingAtty. Dkt. No. 155529.00033

[0522] down onto the solution to react the activated hydrogel surface with the HaloTag ligand. The glass capture slide was incubated on ligand overnight at room temperature and subsequently quenched / blocked in SuperBlock from ThermoFisher (37516) for at least 30 minutes at roomtemperature, with rocking. Blocked capture slides were then used for capture of IVTT proteins as described above.

[0523] All antibodies were diluted into PBST-M. Primary antibodies were diluted 1:750 and the glass slides harboring captured HaloTag fusion proteins were exposed to primary antibody for 1.0 hour at room-temperature, with rocking. After rinsing slides in PBST several times, secondary antib...

Claims

Atty. Dkt. No. 155529.00033CLAIMSWe claim:

1. A system for integrated on-chip drug discover}' of peptide- or protein-based drug molecules or their conjugates, the system comprising:a) a nanowell array comprising,a nanowell slide comprising a plurality of wells; anda biosensor surface comprising an array of discrete locations, wherein a plurality of drug molecules is captured to the biosensor surface to generate a plurality7of on-chip drug molecules, wherein each discrete location comprises one drug molecule of the plurality of drug molecules captured to the biosensor surface,wherein the plurality of drug molecules comprises one or more peptide, or protein; wherein at least a portion of the plurality of drug molecules is configured to be generated on the biosensor surface from nucleic acid sequences using in vitro cell-free or cellbased expression system in the plurality of wells; andwherein the plurality of drug molecules is configured to bind covalently to the biosensor surface and form the plurality of on-chip drug molecules;b) a fluidic module configured to deliver one or more analyte or target molecule to the plurality7of on-chip drug molecules under controlled assay conditions, wherein the plurality7of on-chip drug molecules is configured to be regenerated after assaying with the one or more analyte or target molecule, for multiple successive assay and regeneration cycles; andc) a detection or sensing module configured to measure binding interactions between the plurality of on-chip drug molecules and the one or more analyte or target molecule, wherein the system is configured to perform synthesis or presentation of the drug molecules and measurement of binding interactions on the same solid support surface without transferring the drug molecules off the biosensor surface.

2. The system of claim 1, wherein the plurality of drug molecules is a variant library comprising variants of a single peptide or protein.

3. The system of claim 1 or 2, wherein the plurality of drug molecules are antibodies or antigen-binding proteins.Atty. Dkt. No. 155529.000334. The system of any one of claims 1-3. wherein each well is at a temperature of about 10 deg. C to 37 deg. C.

5. The system of any one of claims 1-4, wherein the nucleic acid sequences comprise 1) dual chain constructs comprising a heavy chain construct and a light chain construct and / or 2) single chain constructs at DNA concentration of about 0.1 nM to about 10 nM.

6. The system of claim 5, wherein the nucleic acid sequences are added as a dual chain construct at a ratio of about 1: 1 to about 1:3 molar ratio of heavy chain construct concentration to light chain construct concentration.

7. The system of any one of claims 1-6, wherein the binding interactions comprise mass spectrometer data, kinetic or equilibrium parameters comprising association rate, dissociation rate, equilibrium dissociation constant (KD), a maximum binding response, residence time, expression yield, or combinations thereof8. The system of any one of claims 1-7. wherein the biosensor surface and the plurality of drug molecules are complexed with a detection tag or a ligand for the detection tag to allow the biosensor surface to capture the drug molecule via interaction of the detection tag and the ligand for the detection tag.

9. The system of claim 8, wherein the detection tag and ligand for the detection tag belong to a system selected from the group consisting of a Halotag system, SNAP -tag system, CLIP-tag system, ACP-tag system, and Spytag-Spy catcher system.

10. The system of claim 8 or 9, wherein the detection tag or the ligand for the detection tag is configured to bind to the biosensor surface covalently.

11. The system of any one of claims 1-10, further comprising a data processing module configured to analyze the binding data to perform one or more integrated drug discovery functions selected from affinity ranking, epitope binning, epitope classification, epitope mapping, on-target and off-target screening, polyreactivity screening, thermalAtty. Dkt. No. 155529.00033stability- analysis, aggregation assessment, hydrophobicity or hy drophilicity- evaluation, solubility evaluation, and pH stability evaluation.

12. The system of any one of claims 1-11, further comprising a computational analysis module operatively coupled to the detection or sensing module and configured to transform the interaction data into drug-discovery outputs comprising at least one of affinity ranking, epitope binning competitive binding relationship, epitope classification, epitope mapping, polyreactivity screening, off-target screening, aggregation assessment, thermal stability', or pH stability'.

13. The system of any one of claims 1-12. wherein the system is configured to assess aggregation propensity of the plurality of on-chip drug molecules by detecting changes in binding signal, kinetics, and / or surface behavior.

14. A method for integrated on-chip drug discovery of peptide- or protein-based drug molecules or their conjugates, the method comprising:(a) synthesizing a plurality' of drug molecules and attaching the plurality of drug molecules to a solid support surface at discrete locations on the solid support surface, wherein the plurality of drug molecules are synthesized from nucleic acid sequences in cell-free systems or cell-based systems in sealed wells, wherein the plurality of drug molecules bind covalently to the solid support surface and form a plurality of on-chip drug molecules;(b) contacting the plurality- of on-chip drug molecules on the solid support with one or more target molecules under controlled assay conditions;(c) measuring binding interactions between the drug molecules and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or mass spectrometer data, or combinations thereof; and(d) analyzing the binding data to perform one or more drug discovery operations selected from affinity ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping,wherein the generating of the drug molecules and the measuring of the binding interactions are performed on the same solid support surface in an integrated workflow.Atty. Dkt. No. 155529.0003315. The method of claim 14, wherein the method comprises performing affinity ranking of the plurality of on-chip drug molecules, the affinity ranking of the drug molecules comprising determining one or more binding parameters selected from the group consisting of an equilibrium dissociation constant, an association rate, a dissociation rate, binding free energy, change in binding free energy, residence time, expression yield, or combinations thereof.

16. The method of claim 15, further comprising affinity ranking the drug molecules based on the one or more binding parameters.

17. The method of claim 15, further comprising assessing aggregation propensity of the on-chip drug molecules, the assessing aggregation propensity comprising detecting changes in binding signal, binding kinetics, or surface behavior of the on-chip drug molecules.

18. The method of claim 17, further comprising reducing the aggregation propensity of the on-chip drug molecules.

19. The method of claim 18, wherein the synthesizing step a) is performed in wells prior to attaching to the solid support surface and each well is at a temperature of about 10 deg. C to 37 deg. C.

20. The method of claim 18 or 19, wherein the synthesizing step a) is performed in wells prior to attaching the drug molecule to the solid support surface and wherein the nucleic acid sequences comprise 1) dual chain constructs comprising a heavy chain construct and a light chain construct and / or 2) single chain constructs at DNA concentration of about 0.1 nM to about 10 nM.

21. The method of claim 20, wherein the nucleic acid sequences are added as a dual chain construct at a ratio of about 1: 1 to about 1:3 molar ratio of heavy chain construct concentration to light chain construct concentration.

22. The method of any one of claims 14-21, wherein the solid support surface and the plurality of drug molecules are complexed w ith a detection tag or a ligand for theAtty. Dkt. No. 155529.00033detection tag to allow the solid support surface to attach to the drug molecule via interaction of the detection tag and the ligand for the detection tag.

23. The method of claim 22, wherein the detection tag and ligand for the detection tag belong to a system selected from the group consisting of a Halotag system, SNAP-tag system, CLIP-tag system, ACP-tag system, and Spytag-Spy catcher system.

24. The method of any one of claims 14-23, further comprising regenerating the plurality of on-chip drug molecules on the solid support surface for multiple, successive contacting, measuring, analyzing and regeneration cycles.

25. The method of claim 24, wherein regenerating the on-chip drug molecules comprises contacting the on-chip drug molecules with a regeneration reagent to remove the target molecules without affecting the on-chip drug molecules covalently bound to the solid support surface, and repeating the contacting, measuring and analyzing steps of the method using a second target molecule to interact with the on-chip drug molecules.

26. The method of claim 25, wherein the regeneration reagent comprises a buffer having a pH of about 2.0 to about 3.0.

27. The method of claim 25 or 26, wherein the regenerating step and the contacting, measuring and analyzing steps may be repeated using the on-chip drug molecules and a different target molecule in each repeat of the method using the on-chip drug molecules.

28. The method of claim 27, wherein the regenerating the on-chip drug molecules is repeated 30 times or more.

29. A method of integrated on-chip drug discovery of peptide- or protein-based drug molecules or their conjugates, the method comprising:a) contacting a plurality of on-chip peptide- or protein-based drug molecules attached covalently to a solid support with one or more target molecules, wherein the on-chip drug molecules are soluble and capable of binding their native ligand;Atty. Dkt. No. 155529.00033b) measuring binding interactions between the drug molecules and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or mass spec data, or combinations thereof; andc) analyzing the binding data to perform one or more drug discovery operations selected from affinity' ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mapping.

30. The method of claim 29, further comprising regenerating the on-chip drug molecules by a method further comprising:d) stripping the one or more target molecules from the plurality of drug molecules while leaving the drug molecules covalently bound to the solid support surface;e) repeating steps a) through c) with a second target molecule.

31. The method of claim 30, wherein the regenerating the on-chip drug molecules and repeating the contacting, measuring and analyzing steps may be repeated up to 30 times with the on-chip drug molecules.

32. A method of integrated on-chip drug discovery' of peptide- or protein-based drug molecules or their conjugates, the method comprising:a) contacting a plurality' of on-chip peptide- or protein-based drug molecules attached covalently to a solid support with one or more target molecules;b) measuring binding interactions between the drug molecules and the target molecules on the solid support to obtain binding data comprising kinetic or equilibrium parameters including association rate, dissociation rate, equilibrium dissociation constant (KD), or mass spec data, or combinations thereof;c) analyzing the binding data to perform one or more drug discovery operations selected from affinity ranking of the drug molecules, epitope binning, epitope classification, epitope mapping, or high-resolution epitope mappingd) stripping the one or more target molecules from the plurality of drug molecules while leaving the drug molecules covalently bound to the solid support surface;e) repeating steps a) through c) with a second target molecule.

Images