Multiparametric Discovery and Optimization Platform
Patent Information
- Application Number
- JP2023528634
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2020-11-13
- Filing Date
- 2021-11-13
- Publication Date
- 2024-11-11
AI Technical Summary
Current molecular and strain discovery and optimization processes are limited by low throughput, high costs, and the inability to integrate multiple data inputs, leading to slow and suboptimal outcomes in directed evolution techniques.
A modular, high-throughput platform utilizing concurrent machine learning and large-scale data integration, including empirical and open data, to drive molecular and strain optimization through guided mutations and directed evolution.
Enhances the efficiency and effectiveness of molecular and strain discovery by leveraging integrated data-driven processes, enabling rapid identification of improved polymer variants with desired functionalities.
Smart Images

Figure 00000000_0000_ABST
Abstract
Description
Technical Field
[0001] Cross-reference This application claims the benefit of U.S. Provisional Application No. 63 / 113,571, filed November 13, 2020, which is hereby incorporated by reference in its entirety.
Background Art
[0002] Background Directed evolution is a useful tool for screening libraries of species and selecting for desired functionality in an iterative process. Such processes enable screening of a large number of molecules or strains (e.g., gene variants) for functional characteristics and manipulation of molecules (e.g., proteins) towards functional goals.
[0003] However, current processes for the discovery and optimization of molecules and / or strains are limited in throughput and / or do not utilize multiple data inputs from various sources to predict results or drive the directed evolution process. Additionally, the high costs and long times associated with conventional high-throughput screening technologies typically limit the diversity of conditions under which products can be optimized. As a result, the discovery and optimization of molecules and / or strains can be slow, costly, and suboptimal. The lack of an integrated multi-parametric and data-driven process may further impede the discovery of molecules or strains or prevent the discovery of molecules or strains that require multiple inputs for discovery.
Summary of the Invention
[0004] Summary This specification recognizes the need for a modular, high-throughput platform for molecular discovery, molecular optimization, and species or strain optimization (e.g., using induced mutation, directed evolution, etc.). This disclosure addresses these needs by providing a system and method for using a combination of large-scale data inputs (including open data and publicly sourced data), empirical data, and known conditions into a concurrent machine learning platform to drive molecular discovery, optimization, and functional prediction of biomolecules (e.g., biomolecules such as proteins).
[0005] In one embodiment, a method for producing an improved polymer variant, A method is provided herein that includes (a) acquiring a dataset, comprising (i) sequence data including polymer variant sequence data, and (ii) enrichment data based on the abundance of polymer variant sequences in a sample selected based on polymer variant function, and (b) applying one or more machine learning algorithms to the dataset to design improved polymer variants.
[0006] In some embodiments, the polymer variant sequence includes sequences in i) an unselected sample and ii) a sample selected based on polymer variant function, wherein the sample selected based on polymer variant function and the unselected sample originate from a common sample.
[0007] In some embodiments, the enrichment data is further based on the abundance of polymer variant sequences in unselected samples. In some embodiments, the abundance of polymer variant sequences in samples selected based on polymer variant function includes sequence counts. In some embodiments, the abundance of polymer variant sequences in unselected samples includes sequence counts. In some embodiments, the enrichment data includes an enrichment factor of the first polymer variant sequence, based on i) the abundance of the first polymer variant sequence in polymer variant sequences in unselected samples, and ii) the abundance of the first polymer variant sequence in polymer variant sequences in a subset of samples selected based on polymer variant function. In some embodiments, the enrichment factor is a functional label of the first polymer variant sequence. In some embodiments, the method comprises performing multiple generations of (a) and (b), and using the improved polymer variant of (b) to generate a dataset of (a) in subsequent generations of the multiple generations. In some embodiments, the multiple generations are at least three or four generations.
[0008] In some embodiments, the polymer variant comprises a protein. In some embodiments, the protein comprises an enzyme. In some embodiments, the function comprises the activity of the enzyme. In some embodiments, the improved polymer variant comprises at least 1000 improved polymer variants. In some embodiments, the polymer variant sequence comprises a nucleic acid sequence encoding the polymer variant.
[0009] In some embodiments, the method further includes providing a library of cells containing polymer variants before acquiring a dataset, wherein the polymer variants comprise multiple polymer variants. In some embodiments, the method further includes generating a library of cells containing polymer variants. In some embodiments, the method includes generating a site-saturated nucleic acid molecule (SSL) that encodes proteins, each having 20 naturally occurring amino acids at each position. In some embodiments, the method further includes performing gene shuffling and error-prone polymerase chain reaction (epPCR) on the SSL to generate nucleic acid molecules encoding polymer variants. In some embodiments, the method further includes transforming cells with the nucleic acid molecules encoding polymer variants to generate a library of cells. In some embodiments, the cells comprise bacterial cells. In some embodiments, the method further includes distributing the library of cells to generate multiple partitions. In some embodiments, each partition of the multiple partitions comprises, on average, one or fewer cells. In some embodiments, the method further includes expanding the library of cells in the multiple partitions. In some embodiments, the method further includes expressing polymer variants in the library of cells in the multiple partitions. In some embodiments, the method further includes determining polymer variant function using the library of cells containing polymer variants. In some embodiments, the polymer variant comprises a protein. In some embodiments, the protein comprises an enzyme. In some embodiments, determining the polymer variant function involves using an enzyme substrate and a reaction buffer. In some embodiments, the polymer variant function comprises enzyme activity. In some embodiments, the enzyme activity produces a fluorescent product.
[0010] In some embodiments, the method further includes sorting a subset of a cell library containing polymer variants based on polymer variant function to provide sorted and unsorted samples. In some embodiments, function is the amount of enzyme activity. In some embodiments, sorting includes sorting a subset of the cell library into several groups based on function. In some embodiments, the several groups are at least four groups. In some embodiments, the subset of the cell library contains about 50 to 100 times the number of unique polymer variants expected to be present in the cell library. In some embodiments, the method further includes determining polymer variant sequences using the polymer variant function-sorted samples. In some embodiments, the polymer variant function-sorted samples include a reference sequence. In some embodiments, determining polymer variant sequences using polymer variant function-sorted samples includes purifying nucleic acid molecules from the polymer variant function-sorted samples. In some embodiments, the method further includes purifying nucleic acid molecules from the unsorted samples. In some embodiments, the purification of nucleic acid molecules from a sample sorted based on polymer variant function occurs separately from the purification of nucleic acid molecules from an unsorted sample. In some embodiments, the method further includes sequencing nucleic acid molecules from the sample sorted based on polymer variant function and nucleic acid molecules from the unsorted sample to provide polymer variant sequences. In some embodiments, sequencing includes next-generation sequencing. In some embodiments, the method further includes using the polymer variant sequences to determine the abundance of polymer variant sequences in i) the sample sorted based on polymer variant function and ii) the unsorted sample.In some embodiments, the method further includes determining enrichment data using the abundance of polymer variant sequences in i) a sample sorted based on polymer variant function and ii) an unsorted sample. In some embodiments, the enrichment data includes an enrichment factor. In some embodiments, the method further includes synthesizing the improved polymer variant. In some embodiments, synthesis includes synthesis on an oligonucleotide array. In some embodiments, the method further includes assembling the improved polymer variant into a vector. In some embodiments, the vector is a plasmid. In some embodiments, the method further includes introducing the vector into host cells.
[0011] In some embodiments, the first generation (a) of multiple generations includes performing saturated mutagenesis and epPCR to provide a library of cells, the second generation (a) of multiple generations includes performing epPCR, and the third generation (a) of multiple generations includes performing epPCR. In some embodiments, each generation of multiple generations includes multiple selections.
[0012] In another embodiment, a non-temporary computer-readable medium is provided herein, which, when executed by a computer processor, causes the computer processor to perform the method described in any embodiment described herein.
[0013] In another embodiment, a system comprising a computer processor and a non-temporary computer-readable medium is disclosed herein.
[0014] In yet another embodiment, a system for processing cells, comprising one or more processors and a computer-readable medium, wherein the computer-readable medium includes instructions, when executed by one or more processors, causing the system to perform (a) distributing a population of cells into a plurality of reaction volumes to obtain distributed cells, and (b) assaying the distributed cells for phenotypic purposes to generate a signal for the distributed cells, wherein the coefficient of variation (CV) of the signal for the distributed cells is 11% or less if the population of cells is a homogeneous population of cells, is disclosed herein.
[0015] In some embodiments, the CV is 10% or less. In some embodiments, the CV is 7% or less. In some embodiments, the CV is approximately 5% to approximately 10%.
[0016] In some embodiments, a homogeneous population of cells includes at least 1 million cells. In some embodiments, a homogeneous population of cells includes at least 10,000,000 cells.
[0017] In some embodiments, the reaction volume includes a droplet.
[0018] In some embodiments, the assay includes adding a detection reagent to the distributed cells. In some embodiments, the assay includes controlling the timing of adding the detection reagent to the distributed cells. In some embodiments, the detection reagent includes an enzyme substrate.
[0019] In some embodiments, the assay includes detecting a signal to distributed cells. In some embodiments, the detection includes optical detection.
[0020] In some embodiments, the assay includes controlling the timing of the detection signal to the distributed cells.
[0021] In some embodiments, the phenotype includes enzyme activity.
[0022] In some embodiments, (a) includes controlling the volume size of each of a plurality of reaction volumes.
[0023] In some embodiments, a homogeneous population of cells contains the same genetic material.
[0024] In some embodiments, a homogeneous population of cells contains the same gene of interest having the same sequence.
[0025] In another aspect, a method for generating improved cells, comprising: (a) generating a library of mutagenized cells; (b) partitioning the library of mutagenized cells into a plurality of partitions to obtain partitioned mutagenized cells; (c) screening the partitioned mutagenized cells for a phenotype; (d) based on the screening, using a threshold gate to select improved cells from the partitioned mutagenized cells; (e) using a feedback controller (i) of (b) to maintain the generation frequency of the plurality of partitions, or using a feedback controller (ii) of (d) to adjust the threshold gate; and (f) repeating steps (a)-(d) using the improved cells, is provided herein.
[0026] In some embodiments, the method includes using a feedback controller to maintain the generation frequency of the plurality of partitions. In some embodiments, the plurality of partitions includes a plurality of droplets.
[0027] In some embodiments, a feedback controller adjusts the flow rate to maintain the generation frequency of a plurality of droplets. In some embodiments, the method includes using a feedback controller to adjust a threshold gate. In some embodiments, the feedback controller raises or lowers the threshold gate. The method of claim 76, wherein a feedback controller adjusts a threshold gate to select a target percentage of mutagenized cells to be dispensed. In some embodiments, the feedback controller is a proportional integral derivative (PID) algorithm. In some embodiments, the feedback controller accounts for deviations in mechanical equipment, liquid viscosity, or temperature.
[0028] In some embodiments, generating a library of mutagenized cells includes contacting the cells with a mutagen. In some embodiments, generating a library of mutagenized cells includes transposon mutagenesis, multiplex automated genome engineering (MAGE), genome shuffling, random recombination, or non-homologous end joining. In some embodiments, generating a library of mutagenized cells includes introducing a heterologous genetic element into the cells. In some embodiments, the heterologous genetic element includes a promoter, 5' untranslated region, ribozyme, RNA stability sequence, nucleic acid encoding a secretion peptide, nucleic acid encoding a signal peptide, nucleic acid encoding a fusion protein, nucleic acid encoding a DNA binding domain, nucleic acid encoding a protein-protein binding region, nucleic acid encoding a codon-optimized gene, 3' untranslated region, or terminator.
[0029] In some embodiments, the mutagenized cells include at least 1000 cells, at least 10,000 cells, or at least 1,000,000 cells.
[0030] In some embodiments, the method further includes expanding the library of mutagenized cells prior to (b). In some embodiments, the expansion occurs in liquid culture.
[0031] In some embodiments, the distribution is controlled by a computer-readable medium. In some embodiments, each of the multiple partitions contains a volume of liquid within a predetermined range. In some embodiments, the method further includes imaging each of the multiple partitions prior to (c). In some embodiments, the method further includes ensuring, based on the imaging, the distribution of the multiple partitions containing a desired number of mutagenic cells.
[0032] In some embodiments, the method further includes subjecting multiple partitions to cell proliferation conditions between (b) and (c).
[0033] In some embodiments, providing the above includes controlling the temperature, carbon source, pH, culture medium components, or aeration of a subset of partitions.
[0034] In some embodiments, the method further includes monitoring cell proliferation in multiple partitions subjected to cell proliferation conditions using imaging.
[0035] In some embodiments, the screening involves adding a detection reagent to each of the multiple partitions. In some embodiments, the detection reagent alters the environment of each of the multiple partitions. In some embodiments, the detection reagent includes pH adjusters, ionic strength adjusters, ions, substrates, inhibitors and enzymes, or competitors.
[0036] In some embodiments, a library of mutagenic cells encodes the molecule of interest. In some embodiments, the screening includes detecting the molecule of interest, measuring the activity of the molecule of interest, which may optionally be catalytic activity and which may optionally produce a fluorescent product, determining the amount of the molecule of interest, or determining the function of the molecule of interest.
[0037] In some embodiments, the screening includes performing multiple screenings. In some embodiments, the multiple screenings include applying a more stringent threshold to the desired phenotype in subsequent screenings compared to screenings prior to the multiple screenings.
[0038] In some embodiments, selecting a partition includes performing a spectroscopic measurement.
[0039] In some embodiments, selecting a partition includes applying a threshold.
[0040] In some embodiments, (f) above includes removing mutagenic cells from a selected partition to obtain improved cells. In some embodiments, the method further includes maintaining the viability of mutagenic cells. In some embodiments, the method further includes growing mutagenic cells.
[0041] In some embodiments, the method further includes screening cells for the desired phenotype.
[0042] In some embodiments, the mutagenic cells include bacteria.
[0043] In another embodiment, computer-readable media are also provided herein, which, when executed by a computer processor, includes instructions causing the computer processor to perform steps (b) to (e).
[0044] In another embodiment, a system including a computer processor and a computer-readable medium is disclosed herein.
[0045] In another embodiment, an integrated system for selecting a desired biological variant is disclosed herein, comprising two or more integrated units selected from the group consisting of strain optimization units, molecular optimization units and molecular discovery units, wherein (a) the strain optimization unit comprises a plurality of separate strain optimization reaction vessels, each having a different strain optimization genetic variant or fermentation conditions; (b) the molecular optimization unit comprises a plurality of separate molecular optimization reaction vessels, each having a different target protein variant or application conditions; and (c) the molecular discovery unit comprises a plurality of separate molecular discovery reaction vessels, each having a different discovery molecule or application conditions.
[0046] In some embodiments, two or more units are integrated via a computer-readable medium capable of performing one or more of the following functions: (i) storing datasets from each of the two or more integrated units in a data repository; (ii) setting threshold parameters for data to be delivered to the data repository; and (iii) performing instructions for each of the two or more units to repeat the input diversity screening.
[0047] In some embodiments, two or more units are integrated via a computer-readable medium that includes a machine learning algorithm that combines data from each of the two or more units for a selected desired biological variant. In some embodiments, the machine learning algorithm includes Elastic Net Regression Generalized Linear Model (GLMNET), Support Vector Machine Regression (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Multilayer Perceptron (MLP), or Convolutional Neural Network (CNN).
[0048] In some embodiments, each of the strain optimization reaction vessel, the molecule optimization reaction vessel, and / or the molecule discovery reaction vessel contains, on average, one or fewer cells.
[0049] In some embodiments, each of the strain optimization reaction vessel, the molecule optimization reaction vessel, and / or the molecule discovery reaction vessel contains more than one cell on average. In some embodiments, the average of more than one cell per well is a clonal population. In some embodiments, the average of more than one cell per well includes a diverse cell population.
[0050] In some embodiments, each of the strain optimization reaction vessel, molecular optimization reaction vessel, and / or molecular discovery reaction vessel contains an average of 10 to 100 cells. In some embodiments, the strain optimization unit, molecular optimization unit, and molecular discovery unit are configured to perform single-cell analysis. In some embodiments, the strain optimization unit, molecular optimization unit, or molecular discovery unit is configured to perform multi-cell analysis.
[0051] In some embodiments, the strain optimization unit, the molecular optimization unit, and / or the molecular discovery unit are configured to screen at least 100 cells per minute.
[0052] In some embodiments, each of two or more integrated units includes a consumption unit. In some embodiments, the consumption unit includes a microfluidic device. In some embodiments, the microfluidic device includes one or more functional elements selected from the group including droplet generators, droplet condensers, buffer reservoirs, cell incubators, fluid injectors, droplet splitters, droplet sorters, delay lines, and droplet singulators.
[0053] In some embodiments, the reaction vessel of the strain optimization reaction vessel, molecular optimization reaction vessel, or molecular discovery reaction vessel has a volume of less than 1 microliter. In some embodiments, the reaction vessel of the strain optimization reaction vessel, molecular optimization reaction vessel, and / or molecular discovery reaction vessel is located in a nanowell array. In some embodiments, the reaction vessel of the strain optimization reaction vessel, molecular optimization reaction vessel, and / or molecular discovery reaction vessel is a droplet, a plug, or a well.
[0054] In some embodiments, the integrated system further includes one or more detectors coupled to such units, the detectors being selected from the group consisting of mass spectrometers, Raman spectrometers, microscopes, cell counters, fluorescence microscopes, optical microscopes, flow cytometers, mass spectrometers, fluorescence plate readers, near-infrared (NIR) spectrophotometers, and piezoelectric sensors.
[0055] In some embodiments, the integrated system further includes a reaction vessel moving actuator selected from the group consisting of robotic manipulators, flow generators, acoustic drop generators, optical tweezers, thermal drop-on-demand devices, and piezoelectric drop-on-demand devices.
[0056] In some embodiments, each of the different strain-optimized genetic variants, each of the target protein variants, and / or each of the target discovery molecules are located in different cells.
[0057] In some embodiments, each strain-optimized genetic variant is generated by random mutagenesis (e.g., exposure to mutagens such as ultraviolet, X-ray, and gamma irradiation; chemical mutagens such as ethyl methyl sulfonate, reactive oxygen species, deamination agents, polycyclic aromatic hydrocarbons, alkylating agents (e.g., ethyl nitrosourea, nitrosamines), aromatic amines, alkaloids, bromine, sodium azide, psoralens, benzene, metals (e.g., arsenic, cadmium, chromium, nickel) or biological agents (e.g., transposons, viruses, bacteria)), multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), or CRISPR-guided DNA polymerase. In some embodiments, random mutagenesis is saturated mutagenesis.
[0058] In some embodiments, each target protein variant is generated by random mutagenesis, multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), or CRISPR-guided DNA polymerase. In some embodiments, random mutagenesis is saturated mutagenesis or cyclic substitution.
[0059] In some embodiments, each discovered molecule is a DNA-coding molecule found in a metagenomic library, a DNA-coding library, a library of DNA from one or more microorganisms, or a library generated by a function generator.
[0060] In some embodiments, the strain optimization unit includes a strain optimization library containing multiple different genetic variants of a first strain. In some embodiments, the multiple different genetic variants are classified by metabolic pathway. In some embodiments, the strain optimization library includes phenotypic, sequence, and metabolic pathway data. In some embodiments, the strain optimization unit includes a machine learning algorithm. In some embodiments, the machine learning algorithm identifies genetic variants optimized across multiple fermentation conditions.
[0061] In some embodiments, each of two or more units includes a recursive cycle that includes additional inputs of biological or conditional diversity.
[0062] In some embodiments, multiple separate reaction volumes are provided, at least 10 5 It contains individual separate reaction volumes.
[0063] In some embodiments, the applicable conditions include one or more of the following: pH, substrate, reaction buffer, and temperature.
[0064] In some embodiments, the fermentation conditions include one or more of the following: pH, oxygenation, carbon source, buffer concentration, and temperature. In some embodiments, the strain optimization library includes at least 100 different variants of the first strain. In some embodiments, the strain optimization library is produced by random mutagenesis of the strain (e.g., exposure to mutagens such as ultraviolet, X-ray, and gamma irradiation; chemical mutagens such as ethyl methyl sulfonate, reactive oxygen species, deamination agents, polycyclic aromatic hydrocarbons, alkylating agents (e.g., ethyl nitrosourea, nitrosamines), aromatic amines, alkaloids, bromine, sodium azide, psoralen, benzene, metals (e.g., arsenic, cadmium, chromium, nickel) or biological agents (e.g., transposons, viruses, bacteria)). In some embodiments, the strain optimization library is selected from the group consisting of whole-genome mutagenesis libraries, genome shuffle libraries, targeted genome libraries, and transposon libraries. In some embodiments, the whole-genome mutagenesis library is a whole-genome random mutagenesis library. In some embodiments, the targeted genome library is a promoter swap library. In some embodiments, the target genome library contains or encodes one or more members selected from the group consisting of promoters, 5' untranslated regions, 3' untranslated regions, ribozymes, RNA stability sequences, secreted peptides, fusion proteins, DNA-binding domains, protein-protein-binding regions, codon-optimizing genes, codon-randomization genes, terminators, and non-homologous end joins. In some embodiments, the transposon library is a promoter insertion library.
[0065] In some embodiments, the molecular optimization unit includes a molecular optimization library containing multiple different genetic variants of a first molecule. In some embodiments, the molecular optimization library is selected from the group consisting of a single-gene random mutation library, a site-saturated library, a small-insert metagenomic library, and a large-insert metagenomic library.
[0066] In another embodiment, an integrated system for selecting a desired biological variant is disclosed herein, comprising a biological diversity data stream and a conditional diversity data stream, wherein (a) each of the data streams is independently programmable, (b) the biological diversity data stream comprises data from screening for different genomic variants of a target protein or a strain expressing a target protein, (c) the conditional diversity data stream comprises data from screening for different conditions of the strain and the target protein, and (d) the biological diversity data stream and the conditional diversity data stream are integrated into an optimization unit configured to receive a target parameter, wherein the system is instructed to (i) introduce additional biological diversity or conditional diversity until a target protein or target strain matching the target parameter is identified, or (ii) identify a target protein or target strain matching the target parameter.
[0067] In some embodiments, the optimization unit includes a computer-readable medium capable of performing one or more of the following functions: (i) storing datasets from the biological diversity data stream and the conditional diversity data stream in a data repository; (ii) setting threshold parameters for the data to be delivered to the data repository; and (iii) instructing the optimization unit to repeat the input diversity screening.
[0068] In some embodiments, the optimization unit includes a computer-readable medium containing a machine learning algorithm that integrates a biological diversity data stream and a conditional diversity data stream for the desired selection of biological variants. In some embodiments, the machine learning algorithm performs an Elastic Net Regression Generalized Linear Model (GLMNET), Support Vector Machine Regression (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Multilayer Perceptron (MLP), or Convolutional Neural Network (CNN).
[0069] In some embodiments, the strain optimization unit is configured to screen 100 different genomic variants of the target protein or strains expressing the target protein every minute.
[0070] In some embodiments, the strain optimization unit is configured to screen 100 cells per minute.
[0071] In some embodiments, each of the strain optimization units comprises a consumption unit. In some embodiments, the consumption unit includes a microfluidic device. In some embodiments, the microfluidic device includes one or more functional elements selected from the group including droplet generators, droplet condensers, buffer reservoirs, cell incubators, fluid injectors, droplet splitters, droplet sorters, delay lines, and droplet singulators. In some embodiments, the microfluidic device includes one or more reaction vessels. In some embodiments, the reaction vessels have a volume of less than 1 microliter. In some embodiments, the reaction vessels are located in a nanowell array. In some embodiments, the reaction vessels are droplets, plugs, or wells. In some embodiments, the droplets are water-in-oil droplets. In some embodiments, the droplets are water droplets on a surface. In some embodiments, the reaction vessels contain cells such that there is an average of about one cell per two or more reaction vessels.
[0072] In some embodiments, the integrated system further includes a detector coupled to the optimization unit, the detector being selected from the group consisting of mass spectrometers, Raman spectrometers, microscopes, cell counters, fluorescence microscopes, optical microscopes, flow cytometers, mass spectrometers, fluorescence plate readers, near-infrared (NIR) spectrophotometers, and piezoelectric sensors.
[0073] In some embodiments, the integrated system further includes a reaction vessel moving actuator selected from the group consisting of robotic manipulators, flow generators, acoustic drop generators, optical tweezers, thermal drop-on-demand devices, and piezoelectric drop-on-demand devices.
[0074] In some embodiments, each of the different genomic variants of the target protein is located in a different cell.
[0075] In some embodiments, each of the different genomic variants of a strain is generated by random mutagenesis (e.g., exposure to mutagens such as ultraviolet, X-ray, and gamma irradiation; chemical mutagens such as ethyl methyl sulfonate, reactive oxygen species, deamination agents, polycyclic aromatic hydrocarbons, alkylating agents (e.g., ethyl nitrosourea, nitrosamines), aromatic amines, alkaloids, bromine, sodium azide, psoralens, benzene, metals (e.g., arsenic, cadmium, chromium, nickel) or biological agents (e.g., transposons, viruses, bacteria)), multiple automated genome engineering (MAGE), or traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), or CRISPR-guided DNA polymerase. In some embodiments, random mutagenesis is saturated mutagenesis or cyclic substitution.
[0076] In some embodiments, each of the different genomic variants of the target protein variant is generated by random mutagenesis (e.g., exposure to mutagens such as ultraviolet, X-ray, and gamma irradiation; chemical mutagens such as ethyl methyl sulfonate, reactive oxygen species, deamination agents, polycyclic aromatic hydrocarbons, alkylating agents (e.g., ethyl nitrosourea, nitrosamines), aromatic amines, alkaloids, bromine, sodium azide, psoralens, benzene, metals (e.g., arsenic, cadmium, chromium, nickel) or biological agents (e.g., transposons, viruses, bacteria)), multiple automated genome engineering (MAGE), or traceable genome engineering via CRISPR technology (CREATE). In some embodiments, random mutagenesis is saturated mutagenesis, cyclic substitution.
[0077] In another embodiment, the following methods are provided herein for selecting a final desired genetic variant, which includes combining data from diversity screening of biologically diverse genetic elements, including genetic elements of a target protein and genetic elements of a strain, with data from conditional screening of diverse conditions, including conditions for target protein expression and conditions for target protein performance; and for identifying a final desired genetic variant using a combined dataset.
[0078] In some embodiments, conditional screening is performed on one or more genetic elements that are part of diversity screening.
[0079] In some embodiments, each cell generates a first cell library containing different genetic elements from a first genetic element library. In some embodiments, the method further comprises each cell generating a second cell library containing different genetic elements from a second genetic element library, and the first and second genetic element libraries are screened independently under multiple conditions. In some embodiments, the first and second genetic element libraries are screened using next-generation sequencing (NGS), e.g., sequencing systems from Illumina®, Pacific Biosciences (PacBio®), Oxford Nanopore®, or Life Technologies (Ion Torrent®), to identify enriched mutations under multiple conditions.
[0080] In some embodiments, diversity screening is performed on variants relevant to strain optimization. In some embodiments, strain optimization results in increased yield, improved fermentation conditions, increased viability, and increased tolerance to ion concentrations. In some embodiments, increased yield is due to increased protein expression and / or increased cell proliferation.
[0081] In some embodiments, using a combined dataset to identify the final desired genetic variant further includes (a) identifying one or more intermediate desired genetic variants having the desired trait, (b) obtaining a recombinant library based on the genetic elements of one or more intermediate desired genetic variants having the desired trait, (c) screening the recombinant library by conditional screening under diverse conditions, and (d) identifying the final desired genetic variant, or a recursive cycle of (a) identifying one or more intermediate desired genetic variants having the desired trait.
[0082] In some embodiments, diversity screening is performed on variants related to target protein discovery. In some embodiments, diversity screening is performed on variants related to target protein optimization. In some embodiments, each of the biologically diverse genetic elements is screened in parallel. In some embodiments, each of the biologically diverse genetic elements is screened sequentially.
[0083] In some embodiments, the combined dataset includes data from one or more assays, including fluorescence-generating assays, colorimetric assays, and reporter cell-based assays. In some embodiments, the fluorescence-generating assay is a direct assay, an indirect assay, a binding assay, or an in vivo reporter assay.
[0084] In some embodiments, conditional screening includes adjusting temperature, culture medium, oxygenation, salinity, pH, mixing, reaction time, metal ion concentration, additive concentration, feed rate, or a combination thereof. In some embodiments, conditional screening includes running two or more different conditions in tandem for a first genetic element. In some embodiments, each of the different conditions is run in parallel. In some embodiments, each of the different conditions is run sequentially.
[0085] In some embodiments, the method further includes generating a library of diverse genetic elements. In some embodiments, generation is performed by random mutagenesis, site-directed mutagenesis, site saturation, or genome shuffling. In some embodiments, the method further includes introducing each genetic element from the library of diverse genetic elements into cells or a cell-free system. In some embodiments, the method further includes separating cells with genetic diversity from the library of diverse genetic elements into different reaction vessels such that each reaction vessel has one or fewer cells. In some embodiments, the reaction vessels are droplets, plugs, or wells.
[0086] In some embodiments, the desired genetic variant is related to strain optimization, target molecule optimization, or molecular discovery from environmental DNA or RNA libraries.
[0087] In some embodiments, biological diversity is for strain optimization. In some embodiments, identifying the final desired genetic variant involves measuring one or more properties of the strain, including expression levels, cell viability, cell proliferation, and sensitivity to divalent cation concentrations.
[0088] In some embodiments, the method further includes generating a library of modified strains by site-directed mutagenesis, random mutagenesis (e.g., exposure to mutagens such as ultraviolet, X-ray, or gamma irradiation; chemical mutagens such as ethyl methyl sulfonate, reactive oxygen species, deamination agents, polycyclic aromatic hydrocarbons, alkylating agents (e.g., ethyl nitrosourea, nitrosamines), aromatic amines, alkaloids, bromine, sodium azide, psoralens, benzene, metals (e.g., arsenic, cadmium, chromium, nickel) or biological agents (e.g., transposons, viruses, bacteria)), multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), or CRISPR-guided DNA polymerase. In some embodiments, random mutagenesis is saturated mutagenesis.
[0089] In some embodiments, biological diversity is for target protein optimization. In some embodiments, identifying a desirable genetic variant involves measuring one or more properties of the target protein, including substrate specificity, substrate state, sensitivity to divalent cation concentration, temperature tolerance, catalytic performance, resistance to protease degradation, pH tolerance, sensitivity to inhibitors, cofactor preference, enzyme rate, half-life, activity when immobilized on a support, or sensitivity to gases. In some embodiments, the substrate state is soluble or surface-adsorbed. In some embodiments, diversity screening further includes enriching genetic elements based on one or more properties of the target protein.
[0090] In some embodiments, the genetic elements of the target protein include gene regions, sequence motifs, and protein structure classes. In some embodiments, the gene encoding the target protein is COL1A1 (human type 1 collagen). In some embodiments, the target protein is a structural protein, enzyme, surface receptor protein, peptide hormone, immune system component, or bioactive peptide. In some embodiments, the target protein is an industrial enzyme. In some embodiments, the target protein is a therapeutic protein. In some embodiments, the peptide hormone is insulin or a signaling factor. In some embodiments, the structural protein is collagen or a food-based protein. In some embodiments, the enzyme is lipase or poly(ethylene terephthalate) hydrolase (PETase).
[0091] In some embodiments, the biological diversity is that of a discovery library. In some embodiments, the discovery library is an environmental DNA (eDNA) library.
[0092] In some embodiments, conditional diversity is one or more fermentation conditions. In some embodiments, one or more fermentation conditions include changes in temperature, culture medium, oxygenation, salinity, pH, mixing, feed schedule, carbon source, or any one or a combination thereof. In some embodiments, the method further includes identifying industrial fermentation conditions for producing a desired genetic variant in a volume greater than 100 milliliters.
[0093] In some embodiments, the conditional diversity is one or more applicable conditions. In some embodiments, the applicable conditions include pH, temperature, substrate, reaction buffer, and adhesion to the support.
[0094] In some embodiments, the method further includes identifying a desired genetic variant, including sequencing. In some embodiments, sequencing includes performing PCR. In some embodiments, sequencing includes performing next-generation sequencing (NGS). In some embodiments, identification includes identifying an open reading frame encoding the desired genetic variant. In some embodiments, the method may further include purifying one or more nucleic acid molecules.
[0095] In yet another aspect of the present disclosure, a method for identifying a final optimized protein or final optimized strain that meets performance criteria is provided herein, comprising: (a) (i) data of a plurality of different genetic elements selected from two or more of the following: a) different genetic variants of a strain, b) different genetic variants of a protein, or c) different genes encoding distinct proteins; and (ii) obtaining a dataset containing data of a plurality of different conditions for each of the plurality of different genetic elements; and (b) integrating the datasets to identify or predict a final optimized protein or final optimized strain that meets performance criteria, or identifying additional data to introduce into (a).
[0096] In some embodiments, performance criteria include the expression level of the target protein, the stability of the target protein, the substrate specificity of the target protein, the activity of the target protein, the growth of the strain, the tolerance of the strain to fermentation conditions, the folding of the target protein, the ratio of the target protein to byproducts, or the modification of the target protein.
[0097] In some embodiments, data under multiple conditions include the stability of the target protein, the strain's tolerance to divalent cation concentrations, the substrate specificity of the target protein, temperature tolerance, pH tolerance, expression level, or strain viability.
[0098] In some embodiments, the dataset is applied to a machine learning algorithm, and the additional data in (b) is identified by the machine learning algorithm. In some embodiments, the additional data is a library of genetic elements or additional conditions.
[0099] In some embodiments, the dataset includes data from one or more fluorescence-generating assays or colorimetric assays. In some embodiments, the fluorescence-generating assay is a direct assay, an indirect assay, a binding assay, or an in vivo reporter assay.
[0100] In some embodiments, integrating the datasets to identify the additional data in (b) further includes (i) identifying the intermediate protein or intermediate strain of interest that satisfies a subset of performance criteria, (ii) identifying a plurality of additional variants having at least 70%, 80%, or 90% homology to the intermediate protein or intermediate strain of interest as additional data to be introduced into (a), and (iii) obtaining for each of the plurality of additional variants additional data for a plurality of different conditions to be introduced into (a). In some embodiments, the method further includes obtaining a) the plurality of additional variants individually if they contain fewer than 100 members, or b) the plurality of additional variants at the population level if they contain more than 100 members.
[0101] In some embodiments, the genetic variants of a strain are derived from a library of genetic variants of the strain. In some embodiments, the library of genetic variants of a strain is selected from the group consisting of a whole-genome mutagenesis library, a genome shuffle library, a target genome library, and a transposon library. In some embodiments, the whole-genome mutagenesis library is a whole-genome random mutagenesis library. In some embodiments, the target genome library contains or codes for one or more members selected from the group consisting of promoters, 5' untranslated regions, 3' untranslated regions, ribozymes, RNA stability sequences, secreted peptides, fusion proteins, DNA-binding domains, protein-protein binding regions, codon-optimizing genes, codon-randomization genes, terminators, and non-homologous end joins. In some embodiments, the target genome library is a promoter-swap library. In some embodiments, the transposon library is a promoter-swap library. In some embodiments, the library of genetic variants of a strain is generated by random mutagenesis, multiple automated genome engineering (MAGE), or traceable genome engineering via CRISPR technology (CREATE).
[0102] In some embodiments, the different genetic variants of a protein are a library of genetic variants encoding the target protein. In some embodiments, the dataset includes genetic variants encoding different proteins, and additional data includes a partial sequence library of variants encoding residues. In some embodiments, the partial sequence library is generated by gene shuffling. In some embodiments, the library of genetic variants encoding the target protein is generated by random mutagenesis, multiple automated genome engineering (MAGE), or traceable genome engineering via CRISPR technology (CREATE). In some embodiments, the library of genetic variants encoding the target protein is selected from the group consisting of single-gene random mutation libraries, site-saturated libraries, small insert metagenomic libraries, and large insert metagenomic libraries.
[0103] In some embodiments, the distinct genes encoding different proteins are an environmental DNA (eDNA) library.
[0104] Another aspect of this disclosure provides a non-temporary computer-readable medium containing machine-executable code that, when executed by one or more computer processors, carries out any of the methods described above or elsewhere in this specification.
[0105] Another aspect of this disclosure provides a system comprising one or more computer processors and computer memory coupled thereto. The computer memory contains machine-executable code that, when executed by one or more computer processors, carries out any of the methods described above or elsewhere in this specification.
[0106] Further aspects and advantages of the present disclosure will be readily apparent to those skilled in the art from the following detailed description, which shows and describes only exemplary embodiments of the present disclosure. As will be understood, other different embodiments of the present disclosure are possible, and some of their details can be modified in various obvious ways, all without departing from the present disclosure. Accordingly, the drawings and description should be considered illustrative and not limiting in nature.
[0107] Embedding by reference All publications, patents, and patent applications referenced herein are incorporated by reference to the same extent as each individual publication, patent, or patent application is specifically and individually indicated as being incorporated by reference. To the extent that any publications and patents or patent applications incorporated by reference conflict with any disclosures contained herein, this specification is intended to supersede and / or take precedence over such conflicting material. [Brief explanation of the drawing]
[0108] Novel features of the present invention are described in detail in the appended claims. A better understanding of the features and advantages of the present invention will be obtained by referring to the following detailed description illustrating exemplary embodiments in which the principles of the present invention are utilized, and to the appended drawings (hereinafter also "Figure" and "FIG.").
[0109] [Figure 1] Figure 1 schematically illustrates an exemplary integrated system including a strain optimization unit, a molecular optimization unit, and a molecular discovery unit for selecting desirable biological variants.
[0110] [Figure 2] Figure 2 schematically shows an exemplary integrated system configured to take multiple parameters as input and output predictions about optimized strains, optimized molecules, or novel molecules with desired functional properties.
[0111] [Figure 3] Figure 3 shows a computer system configured by program or other means to carry out the methods provided herein.
[0112] [Figure 4] Figure 4 shows an exemplary workflow for stock optimization.
[0113] [Figure 5] Figure 5 shows an exemplary workflow for strain optimization, including multiple enrichments.
[0114] [Figure 6A-6C] Figures 6A to 6C show exemplary data obtained from strain optimization. Figure 6A shows the evolution of the cell library over multiple enrichments in the first generation. Figure 6B shows the evolution of the cell library over multiple enrichments in the second generation. Figure 6C shows the evolution of the cell library over multiple enrichments in the third generation.
[0115] [Figure 7] Figure 7 shows a plot of the amount of the target molecule over generations.
[0116] [Figure 8] Figure 8 shows exemplary data obtained from strain optimization across multiple enrichments within a generation.
[0117] [Figure 9] Figure 9 shows exemplary data obtained from strain optimization across multiple enrichment cycles.
[0118] [Figure 10] Figure 10 shows an exemplary workflow for molecular discovery.
[0119] [Figure 11] Figure 11 shows exemplary data obtained from molecular discoveries across multiple enrichments within a generation.
[0120] [Figure 12] Figure 12 shows an example of mutation data discovered from molecular discovery.
[0121] [Figure 13-1] Figure 13 shows exemplary data demonstrating how the functional performance of molecules can be improved by using molecular discovery processes. [Figure 13-2] Figure 13 shows exemplary data demonstrating how the functional performance of molecules can be improved by using molecular discovery processes.
[0122] [Figure 14] Figure 14 shows an exemplary workflow for molecular optimization.
[0123] [Figure 15] Figure 15 shows exemplary data obtained from molecular optimization across multiple enrichments within a generation.
[0124] [Figure 16] Figure 16 shows exemplary data obtained from an intergenerational molecular optimization process.
[0125] [Figure 17] Figure 17 shows exemplary data of an algorithm for preventing or considering drift in the system described herein. [Modes for carrying out the invention]
[0126] Detailed explanation This specification discloses methods and systems for selecting and identifying desired biological variants on a high-throughput platform. Generally, the systems and methods described herein involve using one or more integrated units selected from the group consisting of strain optimization units, molecular optimization units, and molecular discovery units. The methods and systems used herein can utilize large datasets input to or output from one or more of the integrated units (e.g., strain optimization units, molecular optimization units, and / or molecular discovery units) to generate predictions or outputs regarding novel molecules having desired properties, optimized molecules (e.g., having improvements to the desired properties), or strains (e.g., naturally occurring or engineered prokaryotes or eukaryotes) exhibiting such desired properties. In some aspects of the present invention, the platforms, systems, and methods described herein can be applied to directed evolution, molecular discovery, and molecular engineering for optimizing desired properties.
[0127] The following terms are given to illustrate the meaning of the terms used herein, in addition to the understanding of these terms by those skilled in the art. Whenever the terms “at least,” “greater than,” or “and more than,” precede a first number in a set of numbers greater than or equal to two, the terms “at least,” “greater than,” or “and more than,” apply to each number in that set of numbers. For example, 1, 2, or 3 or more is equal to 1 or more, 2 or more, or 3 or more.
[0128] Whenever the terms “less than or equal to,” “less than,” or “and less than” precede the first number in a set of numbers greater than or equal to two, the terms “less than,” “less than,” or “and less than” apply to each number in that set of numbers. For example, 3, 2, or 1 or less is equal to 3 or less, 2 or less, or 1 or less.
[0129] As used herein, the term “biological diversity” generally refers to variations or differences between one or more biological molecules. Biological molecules may include, but are not limited to, ions, small molecules, macromolecules (e.g., peptides, proteins, lipids, carbohydrates, nucleic acid molecules), metabolites, cells, tissues, and / or combinations thereof, molecules that exist in biology or can be synthesized. Variations or differences between one or more biological molecules may be functional, structural, or both. Variations or differences between one or more biological molecules may include biological, chemical, physical, mechanical, optical, thermal, or other differences or variations in activity, function, structure, or properties. Biological diversity may refer to differences in biological molecules that are naturally occurring, found in pathological conditions, or that are not naturally occurring, such as synthetic or engineered biological molecules. Biological diversity may refer to differences in genotype or phenotypic characteristics.
[0130] As used herein, the term “biological variant” generally refers to a form or version of a biological molecule that exhibits variation or diversity. A biological variant may be a member of a group (e.g., a library) that includes different biological molecules such as cells, organelles, nucleic acid molecules, peptides or proteins, lipids, carbohydrates, metabolites, or combinations thereof. Different biological molecules may be of the same type or different types. For example, two biological variants may include two DNA molecules encoding the same protein or peptide. Alternatively, two biological variants may include two DNA molecules encoding different proteins or peptides. In some examples, two or more biological variants may exhibit sequence homology. For example, two or more biological variants may exhibit at least 10%, at least 20%, at least 30%, at least 40%, at least 50%, at least 60%, at least 70%, at least 80%, at least 90%, at least 95%, or more sequence homology or structural homology. In some examples, two or more biological variants may exhibit slight differences in composition, structure, or function. For example, two or more biological variants may differ in composition, structure, or function by less than 1%, less than 0.1%, less than 0.01%, less than 0.001%, less than 0.0001%, or less than that. In some examples, biological variants may be genetic variants or molecular variants.
[0131] As used herein, “genetic variant” generally refers to a member of a group (e.g., a library) containing different genetic elements, such as sequences of RNA, DNA, or other nucleic acid molecules, which may encode the same or different proteins, peptides, motifs, amino acids, or other protein structural elements. Genetic variants may have different nucleic acid sequences, structures, modifications (e.g., methylation), base substitutions, or other differences in genetic elements. Genetic variants may include mutagenic cells. Genetic variants may encode proteins or peptides, or they may be non-coding nucleic acid molecules (e.g., introns, transfer RNA, ribosomal RNA, regulatory RNA, etc.). In some examples, two or more genetic variants may exhibit sequence homology. For example, two or more genetic variants may exhibit sequence homology of at least 10%, at least 20%, at least 30%, at least 40%, at least 50%, at least 60%, at least 70%, at least 80%, at least 90%, at least 95%, or more.
[0132] As used herein, the term “molecular variant” generally refers to a member of a group of different molecules (e.g., a library) including small molecules, therapeutic molecules, polymers or polymer molecules, biological molecules such as RNA, DNA or other nucleic acid molecules, proteins, peptides, motifs, amino acids or other protein structural elements, lipids, carbohydrates, glycoproteins, lipoproteins, or combinations thereof. Molecular variants (e.g., polymer variants) may have different chemical compositions, structures, modifications, charges, masses, or other physical or chemical properties. For example, a polymer variant may be a nucleic acid molecule or a peptide or protein. A nucleic acid molecule may be RNA or DNA. RNA may be coding or non-coding and may include, for example, messenger RNA (mRNA), ribosomal RNA (rRNA), or transfer RNA (tRNA). RNA may be a transcript. RNA may be a small RNA with a length of less than 200 nucleic acid bases, or a large RNA with a length of more than 200 nucleic acid bases. Examples of small RNAs include 5,8S ribosomal RNA (rRNA), 5S rRNA, transfer RNA (tRNA), microRNA (miRNA), small interfering RNA (siRNA), small nucleolar RNA (snoRNA), Piwi-binding RNA (piRNA), tRNA-derived small RNA (tsRNA), and small rRNA-derived RNA (srRNA). RNA may be double-stranded RNA or single-stranded RNA. RNA may be circular RNA. In some examples, two or more molecular variants may exhibit sequence homology or composition homology. For example, two or more molecular variants may exhibit composition homology of at least 10%, at least 20%, at least 30%, at least 40%, at least 50%, at least 60%, at least 70%, at least 80%, at least 90%, at least 95%, or more. In some examples, a biological variant or genetic variant is a molecular variant or contains molecular variants.A molecular variant may be a biological variant or a genetic variant, or conversely, a biological variant may be a genetic variant or a molecular variant.
[0133] Strain optimization, molecular optimization, and molecular discovery This specification discloses systems and methods for selecting and identifying desired biological variants. The systems and methods disclosed herein may generally involve inputting data into an integrated system and outputting data and / or predictions regarding the structure-function relationships of molecules, novel molecules with desired properties, optimized (e.g., manipulated or evolved) molecules with desired properties, strain information (e.g., nucleic acid or protein information of an organism producing a molecule with desired properties), and / or optimized strains with desired performance (e.g., cells, nucleic acid sequences, or protein sequences). The system may be an integrated system comprising one or more integrated units selected from the group consisting of strain optimization units, molecular optimization units, and molecular discovery units. The integrated units may be configured individually or collectively to acquire experimental data or to use experimental data as input along with other inputs, as described herein. Alternatively or additionally, integrated units may be configured individually or collectively to acquire or use large datasets (e.g., from databases, publicly available sources, etc.) as input. Overall, the systems and methods described herein are useful for high-throughput screening of biological variants in platform technologies, using both empirical data and large datasets (e.g., from publicly available databases) for molecular and strain optimization and discovery.
[0134] In some examples, each of one or more integration units takes a dataset (e.g., from a database, empirical data, data generated from other integration units, etc.) as input and outputs data and / or one or more predictions for (i) optimal strains (e.g., using a strain optimization unit that can output information about at least one organism having a coding molecule (e.g., a nucleic acid molecule such as DNA or RNA) for a desired functional trait (e.g., protein or enzyme activity, molecular binding or affinity, or other metabolic or phenotypic traits)), (ii) optimal molecules (e.g., using a molecule optimization unit that can output small molecules, peptides, or proteins), or (iii) novel molecules having a desired functional trait (e.g., using a molecule discovery unit that can output small molecules, peptides, or proteins). In some cases, at least one unit of the integration units is configured to obtain experimental data, which, in some examples, is used as input to the same or different integration units to generate predictions for optimal strains, optimal molecules, or novel molecules.
[0135] (i) Stock Optimization Unit In some examples, the systems and methods disclosed herein utilize a strain optimization unit, which may be useful for screening and optimizing strains (e.g., cells having cell type, genetic variants, etc.) for functional properties in a highly predictable, high-throughput format. Such a strain optimization unit may include the use of one or more feedback loops or controllers, which may help to enhance the predictability and consistency of strain optimization (e.g., predictable functional properties such as the amount of product produced, the activity level of the product, or the consistency of cell generation to be further improved).
[0136] A strain optimization unit may be useful for iteratively screening strains (e.g., eukaryotes or prokaryotes with genetic diversity) for desired functional characteristics. For example, a method for strain optimization may include (a) introducing a library of genetic variants into multiple cells; (b) subjecting the multiple cells having the genetic variants to conditions sufficient to screen multiple cells for desired functional characteristics (e.g., phenotypic traits (e.g., increased protein production or secretion) or metabolic traits); (c) isolating cells having the desired characteristics; (d) generating an additional library of genetic variants, wherein the additional library of genetic variants includes variants encoding the desired characteristics; and (e) repeating (a) to (d) until performance requirements are met. In another example, strain optimization may be performed to generate cells with improved characteristics. In such examples, a method for generating improved cells or cells having improved properties may include: (a) generating a library of mutagenic cells; (b) distributing the library of mutagenic cells into multiple partitions (e.g., reaction vessels) to obtain distributed mutagenic cells; (c) screening the distributed mutagenic cells for phenotype; and (d) selecting improved cells from the distributed mutagenic cells (e.g., using a threshold gate) based on the screening; (e) using a feedback controller to (i) maintain the generation frequency of the multiple partitions in (b), or (ii) adjust the threshold gate in (d); and (f) repeating (a) to (d) using the improved cells.In general, strain optimization units may be useful in generating libraries of genetic variants, introducing genetic variants into cells, culturing cells under various growth, culture, or fermentation conditions as needed, screening cells for desired traits, combining a subset of desired genetic variants, or enriching traits to generate new libraries of genetic variants (e.g., by generating recombinant or transgenes containing the desired traits), and repeating the process until performance criteria or requirements are met (see, for example, Examples 1, 4, and 5).
[0137] A strain optimization unit may include an experimental module, which may include multiple separate strain optimization reaction vessels, each having a different strain optimization genetic variant or fermentation conditions. For example, a strain optimization unit can be used in directed evolution methods and the manipulation of cells or cell populations to have desired functional properties. In such an example, a biological variant such as multiple cells with genetic diversity (which may exist naturally or may be induced, for example, through mutagenesis or transformation, transduction, etc., of transgenes or recombinant genes) or a library of mutagenic cells can be screened in the reaction vessel of the strain optimization unit under multiple fermentation conditions as needed, and cells with the desired functional properties can be enriched (e.g., isolated, collected) and analyzed (e.g., via sequencing) to identify genetic variants (e.g., using nucleic acids or gene sequences). Genetic variants possessing desired characteristics (e.g., a target gene, protein expression under specified fermentation conditions, or other phenotypic or metabolic traits) can be isolated and re-screened by recombining them with multiple recombinant genes, e.g., by transformation, transfection, or transduction of recombinant genes into cells, cell culture as needed, and enrichment or isolation of cells possessing desired functional characteristics and, if necessary, more stringent performance requirements. The process can be repeated in recursive cycles so that each cycle (also referred herein as “diversity screening”) has more stringent performance requirements (e.g., higher or lower metabolic activity, enzyme activity, or higher or lower expression of phenotypic traits, increased yield under a series of fermentation conditions, robustness, such as tolerance to differences in ion concentration, temperature, pH, etc., increased viability, increased cell proliferation, or increased growth). Alternatively or additionally, multiple processes can be performed in parallel and / or sequentially. Recursive cycles and the operations contained therein may be performed over as many cycles as necessary to satisfy the desired performance requirements.In some cases, additional inputs of biological diversity (e.g., through gene recombination and introduction of recombinant genes into cells, or through mutagenesis, e.g., exposure to mutagenic substances such as ultraviolet light or ethyl methyl sulfonate) may be introduced during the recursive cycle. Alternatively or additionally, variability of conditions (e.g., fermentation or changes in culture conditions) may be introduced during the recursive cycle.
[0138] In some embodiments, strain optimization can be achieved by performing multiple screenings and enrichments to generate new strain generations by adding biological diversity (e.g., through mutagenesis), and then performing additional multiple screenings and enrichments. In one example, a library of mutagenetic cells can be generated and cultured under conditions sufficient to grow the cells as needed. The library of mutagenetic cells can then be distributed into reaction vessels and cultured under conditions sufficient to grow or expand the cells as needed (e.g., in liquid culture, in a specified set of fermentation conditions). In some examples, individual cells can be distributed into individual reaction vessels. Such distribution can be assisted by computer-readable media and feedback control, as described elsewhere in this specification. Subsequently, the distributed cells can be subjected to screening, and based on the screening, cells having the desired functional properties (e.g., desired phenotype) can be enriched (e.g., isolated, collected, or selected). These enriched cells can then be subjected to additional multiple enrichments (e.g., redistribution, culture, screening, and enrichment). After enrichment a predetermined number of times (e.g., user-determined or computer-induced), the enriched cells can be further analyzed (e.g., via sequencing) to identify genetic variants. The most desirable genetic variant (e.g., one that yields the highest functional performance) can be isolated and recombined with multiple recombinant genes to provide a new generation of biological variants (e.g., genetic variants, mutagenic cells, etc.). The new generation of biological variants can then be rescreened by culturing the cells as needed, for example, by transformation, transfection, or transduction of recombinant genes, and enriching or isolating cells that have the desired functional characteristics and, if necessary, more stringent performance requirements. See, for example, Figures 4-5.
[0139] To achieve the desired functional performance, any number of rounds or generations can be repeated. For example, within a single generation, enrichment or selection may be performed at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, 500, 1,000, or more times. Alternatively, within a single generation, enrichment or selection may be performed up to approximately 1,000, 500, 100, 50, 100, 9, 8, 7, 6, 5, or less times. Similarly, any number of generations (e.g., mutagenesis or addition of biological diversity) can be performed, and for example, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 50, 100, or more generations can be generated for optimization. Alternatively, up to approximately 100 generations, up to approximately 50 generations, up to approximately 10 generations, up to approximately 9 generations, up to approximately 8 generations, up to approximately 7 generations, up to approximately 6 generations, up to approximately 5 generations, up to approximately 4 generations, up to approximately 3 generations, up to approximately 2 generations, or fewer generations (e.g., multiple mutations).
[0140] In some cases, it may be useful to use a strain optimization unit to screen various fermentation conditions. In such cases, multiple cells, including, as necessary, genetic diversity (which may be naturally occurring or induced, as described herein), can be cultured under different fermentation conditions. Non-limiting examples of different manipulable parameters of fermentation conditions include temperature, feed rate, growth or nutrient medium composition, oxygenation, salinity, pH, carbon source, carbon dioxide concentration, buffer concentration, ion concentration, culture duration, perfusion or mixing, aeration, reaction time, metal ion concentration, additive concentration, feed schedule, combinations thereof, or changes to any one or a combination thereof. One or more manipulable parameters of fermentation conditions can be adjusted for a given set of genetic variants. Multiple cells can be cultured under specified fermentation conditions before distributing the cells into reaction vessels, or the cells can be cultured under specified fermentation conditions within the reaction vessels. Fermentation conditions may be the same or different across a subset or all of the reaction vessels, and may be screened independently, sequentially, or in parallel. Screening such variants under various fermentation conditions can be useful in identifying genetic variants that result in functional properties under specific fermentation conditions, such as increased yield, growth, or protein expression levels in environments like low oxygen, low or high temperatures, abnormal pH, or low glucose. In some cases, screening variants under various fermentation conditions can be useful in identifying industrial fermentation conditions (e.g., in volumes greater than 100 milliliters (mL), 1 liter (L), 2 L, 5 L, or more) that can produce desired functional properties.
[0141] Next, cultured cells can be screened in the reaction vessel of the cell line optimization unit, and cells possessing the desired functional properties under specified fermentation conditions can be isolated, collected, and analyzed (e.g., via sequencing) to identify genetic variants (e.g., using nucleic acids or gene sequences). In some cases, sequencing can be used to identify mutations enriched under specified fermentation or culture conditions. Genetic variants possessing the desired properties (e.g., the gene of interest, protein expression under specified fermentation conditions, or other phenotypic or metabotropic traits) can be isolated and, if necessary, further cultured (e.g., under conditions sufficient to maximize the viability or growth of the organism), recombined with multiple recombinant genes, e.g., transformation, transfection, or transduction of recombinant genes into cells, culture of cells under fermentation conditions, and rescreened by isolating cells with the desired functional properties and, if necessary, with more stringent performance requirements (e.g., more stringent or growth under specified fermentation conditions). The process can be repeated in a recursive cycle so that each cycle has more stringent performance requirements (e.g., more stringent fermentation conditions, e.g., hypoxic conditions, high or low temperature, low glucose, etc.). The recursive cycle and the operations contained therein may be performed for as many cycles as necessary to satisfy the desired performance requirements. In some examples, multiple processes can be performed in parallel. In some examples, additional inputs of biological diversity (e.g., via gene recombination and introduction of recombinant genes into cells, or via mutagenesis, e.g., via exposure to mutagenic substances such as ultraviolet light or ethyl methyl sulfonate) can be introduced during the recursive cycle to generate new generations of strains or mutagenic cells. Alternatively or additionally, variability of conditions (e.g., changes in fermentation or culture conditions) may be introduced during the recursive cycle.
[0142] Cells can be cultured in any useful or user-friendly process. For example, cells can be cultured after introducing biodiversity (e.g., via targeted or untargeted mutagenesis). Cells can be cultured after distribution, screening, concentration, or after additional cycles that add biodiversity. It will be understood that cells can be cultured under different conditions at different stages depending on the process. For example, cells can be cultured under suboptimal conditions (e.g., hypoxia, low nutrients) before screening, but after concentration, cells can be cultured under more optimal fermentation conditions to promote growth.
[0143] Using the systems and methods described herein, multiple experimental conditions can be screened simultaneously. For example, the strain optimization unit can be configured to simultaneously screen or analyze millions of reaction volumes or biological variants (e.g., cells with hundreds to millions or more genetic variants) grown under the same or different fermentation conditions. Screening may be carried out by performing one or more assays described herein, for example, through imaging of reaction vessels and sorting of cells and / or reaction vessels based on the imaging results. Beneficially, the strain optimization unit can provide multiplexed multiparametric information regarding the environmental (or fermentation) conditions of genetic and biological variants. Furthermore, the strain optimization unit may be useful for selecting strains based on desired performance criteria (e.g., having desired characteristics such as threshold enzyme activity levels or other metabolic or phenotypic traits) or fermentation conditions.
[0144] Reaction vessels: The reaction vessels of the strain optimization unit may be, for example, droplets or plugs, wells (e.g., microwells, nanowells, picowells, etc.), vials, tubes, flasks, or other containers. In some examples, the reaction vessels are contained within a microfluidic device. In some examples, the reaction vessels are members of an array, such as a microwell array or plate (or other substrate), a nanowell array or plate (or other substrate). Multiple types of reaction vessels (e.g., droplets and wells) may be used simultaneously or sequentially for any of the processes described herein. The reaction vessels may be physically separate or distinct, or they may be able to merge or combine. The reaction vessels may contain substrates on which one or more reagents can be adsorbed, or the reagents may be in solution.
[0145] Each reaction vessel may contain different strain-optimized genetic variants for screening. For example, it may be useful to screen a library of genetic variants for a desired functional property (e.g., biomolecular (e.g., enzyme) activity, biomolecular interactions, phenotype, or metabolic trait), and therefore each reaction vessel may be configured to contain different genetic variants for screening. In some cases, each reaction vessel contains one or more cells having a genetic variant. For example, each reaction vessel may contain a single cell or multiple cells having a single genetic variant that is different from the genetic variants of other cells in other reaction vessels. Alternatively or additionally, each reaction vessel may contain a clonal population of cells having a single genetic variant. In other examples, it may be useful to screen different genetic variants within a single reaction vessel, and therefore the reaction vessel may contain a mixture of cells having different genetic variants. In such examples, each reaction vessel may contain combinations of cells and genetic variants. In some examples, such as with droplets, one or more droplets can be combined or merged to combine cells or genetic variants.
[0146] A reaction vessel may contain a single cell or multiple cells. For example, a single reaction vessel may contain 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 50, 100, 200, 500, 1,000 or more cells. One or more reaction vessels may contain different numbers of cells; for example, a reaction vessel may contain an average of 10 to 100 cells. One or more cells may have the same origin or different origins. One or more cells may be of the same type or different types. One or more cells may originate from any useful species or origin, such as bacteria, yeast, mammals (e.g., humans, primates, rats, mice, rabbits, goats, pigs, cattle, sheep, horses), reptiles, birds, etc. One or more cells may be prokaryotes, eukaryotes, or a combination thereof. Cells may originate from unicellular or multicellular organisms. Cells may originate from cell samples obtained from subjects. Subjects may include mammals (e.g., humans, primates, rats, mice, rabbits, goats, pigs, cattle, sheep, horses), reptiles, birds, etc. Cell samples can be obtained by biopsy, swab, irrigation, brushing, collection, etc. Each reaction vessel may be configured to perform analysis of a single cell or multiple cells.
[0147] In some examples, the reaction vessel does not need to contain cells, but may contain reaction conditions sufficient to carry out cellular processes (e.g., transcription, translation, enzymatic reactions, etc.). In such cases, the reaction vessel may include a cell-free system containing any useful reagents for carrying out such cellular processes, including but not limited to enzymes (e.g., DNA polymerase, RNA polymerase, reverse transcriptase, helicase, ligase), nucleic acid molecules (e.g., nucleotides, transfer RNA), proteins (e.g., ribosomes), lipids, carbohydrates, metabolites, or any combination thereof.
[0148] Reaction vessels can be configured to be evaluated or analyzed in a high-throughput manner. For example, a strain optimization unit (e.g., an experimental module) can be configured to screen at least 100 cells / min, at least 500 cells / min, at least 1,000 cells / min, at least 5,000 cells / min, at least 10,000 cells / min, at least 50,000 cells / min, at least 100,000 cells / min, or more. Alternatively or additionally, reaction vessels can be screened in a high-throughput manner. For example, a strain optimization unit can be configured to screen at least 10 reaction vessels / min, at least 100 reaction vessels / min, at least 500 reaction vessels / min, at least 1,000 reaction vessels / min, at least 5,000 reaction vessels / min, at least 10,000 reaction vessels / min, at least 50,000 reaction vessels / min, or at least 100,000 reaction vessels / min. In some examples, the strain optimization unit can be configured to screen at least 10 reaction volumes / min, at least 100 reaction volumes / min, at least 500 reaction volumes / min, at least 1,000 reaction volumes / min, at least 5,000 reaction volumes / min, at least 10,000 reaction volumes / min, at least 50,000 reaction volumes / min, and at least 100,000 reaction volumes / min.
[0149] Any number of reaction vessels can be generated or used within a given period. In some examples, it may be beneficial to screen different numbers of reaction vessels (or volumes). Therefore, the number of reaction vessels (or volumes) used or generated can be adjusted according to the appropriate application or experimental objective. For example, if a large number of separate reaction volumes are useful for a screening purpose (e.g., for strain optimization), then at least 10,000 reaction volumes, at least 100,000 reaction volumes, at least 1,000,000 reaction volumes, at least 10,000,000 reaction volumes, at least 100,000,000 reaction volumes, or more reaction volumes may be used. In other examples, a smaller number of separate reaction volumes may be useful or required (e.g., for targeted screening, or for the rational design of an experiment using a limited number of species for screening). In such cases, reaction volumes of up to 100,000,000, 10,000,000, 1,000,000, 100,000, 10,000, or less may be used or produced.
[0150] In some examples, the reaction vessel is contained within a microfluidic device. The microfluidic device may be part of a consumable or disposable unit of a strain optimization unit. In some cases, the reaction vessel may include nanowells or droplets (or plugs). In such cases, the microfluidic device may include one or more functional elements such as droplet generators, droplet condensers, droplet conjugaters, buffer reservoirs, cell incubators, fluid injectors, droplet splitters, droplet sorters, delay lines, and droplet singulators.
[0151] The reaction vessel can accommodate any useful volume. The reaction vessel may have a volume of less than 100 microliters (μL), less than 10 μL, less than 1 μL, less than 500 nanoliters (nL), less than 200 nL, less than 100 nL, less than 50 nL, less than 20 nL, less than 10 nL, less than 5 nL, less than 2 nL, less than 1 nL, less than 500 picoliters (pL), less than 200 pL, less than 100 pL, less than 50 pL, less than 20 pL, less than 10 pL, less than 5 pL, less than 2 pL, less than 1 pL, less than 500 femtoliters (fL), or less than these. In some embodiments, the reaction vessel may have a volume greater than 500 fL, greater than 1 pL, greater than 2 pL, greater than 5 pL, greater than 10 pL, greater than 20 pL, greater than 50 pL, greater than 100 pL, greater than 200 pL, greater than 500 pL, greater than 1 nL, greater than 2 nL, greater than 5 nL, greater than 10 nL, greater than 20 nL, greater than 50 nL, greater than 100 nL, greater than 200 nL, greater than 500 nL, greater than 1 μL, greater than 2 μL, greater than 5 μL, greater than 10 μL, greater than 100 μL, or greater than these. It will be understood that the reaction volume may be within a range of volumes such as about 1 pL to about 20 pL, about 10 pL to about 100 nL, etc. Multiple reaction vessels may be used, and the multiple reaction vessels may contain the same or different volumes. As further described elsewhere in this specification, the volumes of the multiple reaction vessels can be controlled using computer implementation methods.
[0152] In the case of droplets, the droplets may contain any useful composition. For example, droplets may include water-in-oil droplets, oil-in-water droplets, water-oil-water droplets, oil-water-oil droplets, and so on. In some examples, droplets may include droplets in the air on or adjacent to a surface (e.g., oil, water) and may be contained in a reaction vessel (e.g., a well). Droplets may contain multiple emulsions, such as double emulsions, triple emulsions, and so on.
[0153] Assays: In some cases, the screening of biological variants may involve the use of one or more screening operations. Screening operations may include assays, e.g., direct assays, indirect assays, binding assays, reporter assays, fluorescence assays, reporter cell-based assays, chromogenic assays, or colorimetric assays. One or more immunological assays may be used, including but not limited to enzyme-linked immunosorbent assays (ELISAs), sandwich ELISAs, and proximity ligation assays. In some cases, a reporter molecule or selection molecule is used for screening. The assay molecule or reporter molecule or selection molecule may be useful for detecting the presence of genetic variants or functional characteristics (e.g., phenotype or metabotropy) of the biological variant. For example, if the desired biological variant is a cell containing enzyme activity, the cells in the reaction vessel may be screened in the presence of a chromogenic or fluorescence-generating substrate that changes color or fluoresces when the substrate is converted to a product by the desired enzyme activity. In such examples, a reaction vessel exhibiting a change in color or fluorescence may indicate that cells possess the desired enzyme activity, and that these cells can subsequently be isolated and further analyzed (e.g., via sequencing such as next-generation sequencing (NGS) or PCR-based sequencing, open reading frame identification or analysis, protein analysis, etc.) to determine information about the biological variant (e.g., DNA or RNA sequence, protein sequence, or composition). In other examples, the desired biological variant may include a selection molecule or a reporter molecule, for example, in the case of a selection molecule, only cells with the desired traits can survive under given culture conditions (e.g., only cells with a transgene or recombinant gene containing antibiotic resistance can survive in a growth medium containing antibiotics). The surviving cells can be screened for the desired functional properties, and those with the desired functional properties can be subjected to further analysis. In yet another example, only cells with the desired traits may express a reporter molecule (e.g., a fluorescent or detectable protein) indicating the presence of the transgene or recombinant gene.In such cases, cells exhibiting desirable traits or functional characteristics can be isolated from the reaction vessel and subjected to further analysis, as described herein. In other cases, assays (e.g., immunoassays, binding assays, etc.) can be used to determine or measure the presence of the target molecule (e.g., protein, nucleic acid molecule, lipid, carbohydrate, metabolite, or combination thereof) in each genetic variant, and the genetic variant containing the target molecule can be isolated and further analyzed.
[0154] Further analysis can be performed on selected biological variants (e.g., selected cells exhibiting desired characteristics). Such analyses include, but are not limited to, protein analysis (e.g., mass spectrometry, protein assays), nucleic acid analysis (e.g., sequencing such as next-generation sequencing (NGS)), lipid analysis (e.g., Raman scattering or spectroscopy), imaging (and image analysis), spectroscopy, or other analytical methods or screenings, as described herein. Such analytical methods may be useful in determining the identity of genetic variants by obtaining information, for example, on DNA sequences or genes, amino acid sequences, or protein properties (e.g., post-translational modifications). Such information can then be collected and stored in a database, or, if necessary, reintegrated with a strain-optimized library having any functional characteristics such as desired biological activity, phenotype, or metabotropy. Such further analysis can be performed in any useful or user-friendly process. For example, in some cases, it may be beneficial to identify or characterize a set of biological variants (e.g., via sequencing or other protein analysis) before determining the functional performance of the biological variants. In such cases, such characterization, e.g., sequencing and optional indexing (of a portion of a clonal population of cells), can be performed before screening biological variants for functional performance. Following the screening for functional performance, sequences may be indexed or paired against functional performance (e.g., using a computer-readable medium), thereby providing information on the correlation between structural (e.g., sequence) information and functional performance.
[0155] In some examples, the strain optimization unit may include or be coupled to a detector. The detector may be used to determine the presence or absence of a target molecule or functional property in the cell or reaction vessel, enabling screening and / or selection of desired biological variants. The detector may include one or more analytical instruments, such as a mass spectrometer, Raman spectrometer, near-infrared (NIR) spectrophotometer, Fourier transform infrared (FTIR) spectrometer, spectrophotometer (e.g., fluorescence or absorbance plate or cuvette reader), piezoelectric sensor, etc. The detector may be configured to count cells (e.g., cell counter, flow cytometer). The detector may include one or more optical components, and in some examples, it may include a microscope (e.g., stereoscopic mirror, compound microscope, fluorescence microscope, polarizing microscope, confocal microscope, differential interference microscope, super-resolution microscope, etc.), or components or combinations thereof. Detectors may be used to detect physical properties of cells or reaction vessels, including but not limited to magnetic properties, impedance, topography or geometric shape, scattering, morphology, absorbance, fluorescence, color, cell clustering, and cell or colony size. Detectors may be point detectors, arrays, or image detectors.
[0156] Moving Actuator: In some examples, a reaction vessel moving actuator is used to move the reaction vessel (e.g., toward a detector). The vessel moving actuator may be a fluid handling unit such as a robotic manipulator, flow generator, acoustic drop generator, optical tweezers, thermal drop on demand, piezoelectric drop on demand, lithography unit, digital microfluidic device, dielectrophoresis, or a combination thereof. The moving actuator may comprise a stage and any necessary moving parts such as screws, gears, threads, motors, pumps, compressors, resistors, transistors, springs, pulleys, etc. In some examples, the reaction vessel is positioned on or adjacent to a substrate connected to the moving actuator.
[0157] Library: A strain optimization unit may include a strain optimization library, which may be used in the processes described herein. The strain optimization library may include a library of biological diversity; for example, the strain optimization library may include multiple different biological variants, such as genetic variants. Multiple genetic variants may originate from a first strain. For example, cells of a first cell line (e.g., bacteria, yeast, animals, plants, etc.) can be subjected to conditions sufficient to produce multiple different genetic variants, such that each genetic variant is distinct. Each genetic variant of the multiple genetic variants may include, for example, DNA or RNA having a different sequence from other genetic variants of the multiple genetic variants. Multiple different genetic variants may encode mutations in the same biomolecule (e.g., DNA, RNA, peptides, proteins, etc.). For example, multiple different genetic variants may include different DNA sequences encoding mutations in RNA sequences that are later translated into proteins. Mutations in different DNA sequences in the library may result in different RNA sequences that may encode the same or different proteins. In some examples, the proteins encoded by genetic variants (which may have mutations in DNA and / or RNA) are enzymes. In such cases, the enzyme may form or be part of a metabolic pathway. The enzyme or metabolic activity may be used as a screening parameter or performance requirement during the strain optimization process.
[0158] In some cases, the strain optimization unit is configured to screen genetic variants and select and / or classify desired biological variants (e.g., genetic variants) based on phenotypic or metabotropic measurements. In non-limiting examples, phenotypic or metabotropic measurements may include protein or enzyme activity, protein expression levels, and quantities of by-products or metabolites. For example, a desired biological variant may have a specified or minimum threshold enzyme activity. In such cases, genetic variants in a strain optimization library can be screened by the strain optimization unit, and only those with the specified or minimum threshold enzyme activity are selected for further analysis. The genetic variants can then be subjected to further analysis (e.g., sequencing, protein analysis) to determine the nucleic acid or protein sequence of the enzyme of interest. The determined sequence information (e.g., nucleic acid or protein sequence) and corresponding functional performance may be input into one or more computer-readable media (e.g., simulators or models), which may then be used to correlate structural elements (e.g., nucleic acid or protein sequences) with functional performance, as described elsewhere in this specification. In some cases, nucleic acid sequences can be recombined with other nucleic acid sequences exhibiting desirable properties (e.g., the same or different enzyme activity), thereby generating a new library of genetic variants. This process can be repeated until the desired threshold of enzyme activity is reached.
[0159] In some cases, strain optimization libraries include generated or synthesized genetic variants. Multiple genetic variants may be generated by one or more techniques. For example, genetic variants can be generated by mutagenesis (random mutagenesis (e.g., exposure to mutagens such as ultraviolet, X-ray, and gamma irradiation, chemical mutagens such as ethyl methyl sulfonate, reactive oxygen species, deamination agents, polycyclic aromatic hydrocarbons, alkylating agents (e.g., ethyl nitrosourea, nitrosamines), aromatic amines, alkaloids, bromine, sodium azide, psoralens, benzene, metals (e.g., arsenic, cadmium, chromium, nickel) or biological agents (e.g., transposons, viruses, bacteria)), saturated mutagenesis, cyclic substitution), multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), CRISPR-guided DNA polymerase, contact with mutagen, performing error-prone polymerase chain reaction (epPCR), or other genome engineering approaches. In some examples, the strain optimization library is or includes targeted genome libraries such as whole-genome mutagenesis libraries, genome shuffle libraries, promoter swap libraries, and transposon libraries (e.g., promoter insertion libraries). In some embodiments, the targeted genome library includes one or more heterogeneous elements. The heterogeneous elements may include or encode one or more members selected from the group consisting of promoters, 5' untranslated regions, 3' untranslated regions, ribozymes, RNA stability sequences, secreted peptides, fusion proteins, DNA-binding domains, protein-protein-binding regions, codon-optimization genes, codon-randomization genes, terminators, and non-homologous end joins. The strain optimization library may include genetic variants generated using targeted mutagenesis or non-targeted (e.g., random) mutagenesis. In some examples, the strain optimization library includes a library of mutagenetic cells.
[0160] A strain optimization library may contain a number of genetic variants useful for screening or optimizing strains of a desired biological variant. A strain optimization library may contain at least 100 genetic variants, at least 1,000 genetic variants, at least 10,000 genetic variants, at least 100,000 genetic variants, at least 1,000,000 genetic variants, at least 10,000,000 genetic variants, at least 100,000,000 genetic variants, or more. As described herein, the strain optimization process may be recursive or iterative. In such cases, the number of genetic variants screened in each iteration may vary; for example, the first cycle may screen 10,000,000 genetic variants, and the second cycle may screen 100,000 genetic variants. Conversely, the number of genetic variants can increase between cycles, for example, through the introduction of new biological diversity or variants (e.g., new strains, new genetic variants, etc.) and / or conditional diversity (e.g., various fermentation conditions for some or all of the newly introduced or screened biological variants).
[0161] Further examples of systems and methods for strain optimization can be found, for example, in J. Agresti et al., 2009, "Ultrahigh-throughput screening in drop-based microfluidics for directed evolution," PNAS, which is incorporated in its entirety herein by reference. See also Examples 1, 4 and 5 below.
[0162] (ii) Molecular Discovery Unit In some examples, the systems and methods disclosed herein utilize molecular discovery units that may be useful for discovering novel molecules (e.g., biomolecules such as proteins and nucleic acids) with functional properties. The systems and methods described herein may include high-throughput screening of a variant library, obtaining information or characterization of the variants, and then using the obtained information or characterization to design novel molecules that are predicted to have functional properties in a high-throughput and consistent manner.
[0163] A molecular discovery unit can be useful for discovering novel molecules with desired functional properties. For example, a molecular discovery unit can be configured to acquire or collect discovery molecules (e.g., DNA from a sample or library, or DNA information from a library) and screen the collected discovery molecules for desired functional properties such as phenotype or metabotropy (e.g., enzyme activity). By isolating molecules exhibiting the desired functional properties and characterizing them (e.g., using sequencing), data on the discovery molecules and their functional properties (e.g., enzyme activity levels, target protein expression, etc.) can be generated. The data may then be input into a computer-readable medium (e.g., a simulator or model) which may include a database, and which can be configured to predict structure-function relationships of discovery molecules, the expected performance of a given structure or sequence under a given set of application conditions, or other predictions based on model inputs. In some embodiments, the computer-readable medium includes machine learning algorithms.
[0164] A molecular discovery unit may include an experimental module, which may include a plurality of separate molecular discovery reaction vessels, each having a different discovery molecule or application conditions. In one example, a method for molecular discovery may include (a) collecting or extracting DNA sequences, (b) introducing the collected or extracted DNA sequences into a plurality of cells, (c) subjecting the plurality of cells having the collected or extracted DNA sequences to conditions sufficient to screen the plurality of cells for a desired functional characteristic (e.g., phenotype or metabotropy), (d) isolating cells having the desired characteristics, and (e) determining the sequences of the isolated cells. The obtained data (e.g., sequencing reads or data) may be input into a computer-readable medium (e.g., a simulator or model), which may be used to predict which sequences, regions, or motifs have the desired functional characteristics. If necessary, the computer-readable medium may be configured to compare the obtained data (e.g., sequencing reads or data) with a database of the same type of data (e.g., known sequences of proteins having the desired characteristics). One or more processes in a molecular discovery unit may be recursive or iterative. Alternatively or additionally, multiple processes may be executed in parallel and / or sequentially.
[0165] In another example, a method for molecular discovery may involve generating improved cells by discovering biological variants (e.g., genetic variants) that result in improved functional performance of the phenotype. Similar to strain optimization, such a method may include, for example, (a) generating a library of mutagenic cells, (b) distributing the library of mutagenic cells into multiple partitions to obtain distributed mutagenic cells, (c) screening the distributed mutagenic cells for phenotype, and (d) selecting improved cells from the distributed mutagenic cells based on the screening, and (e) repeating (a) to (d) using the improved cells. The molecular discovery process may be useful for screening molecular variants (e.g., a library of mutagenic cells) for desired functional properties of the phenotype, analyzing the molecular variants, and determining structure-function relationships based on the observed phenotype and the analysis obtained.
[0166] In one example, a molecular discovery unit may provide the structural-functional relationship of gene sequences to phenotypic traits, such as protein production or enzyme catalytic activity. Such discoveries may be achieved by performing multiple screenings and enriching any additional biological diversity (e.g., via mutagenesis), generating new generations of biological variants, and performing additional screenings and enrichments. For example, a library of mutagenetic cells may be generated and cultured under conditions sufficient to grow the cells as needed. The library of mutagenetic cells may then be distributed into reaction vessels and cultured under conditions sufficient to grow or expand the cells as needed (e.g., in liquid culture, in a specified set of fermentation conditions). In some examples, individual cells may be distributed into individual reaction vessels. Such distribution may be assisted by computer-readable media and feedback control, as described elsewhere in this specification. Subsequently, the distributed cells may be subjected to screening, and based on the screening, cells having the desired functional properties (e.g., desired phenotype) may be enriched (e.g., isolated, collected, or selected). These enriched cells may be subjected to additional enrichments (e.g., redistribution, culture, screening, and enrichment). After enrichment a predetermined number of times (e.g., user-determined or computer-induced), the enriched cells can be further analyzed (e.g., via sequencing) to identify genetic variants (e.g., using nucleic acids or gene sequences).
[0167] In some cases, structure-function relationships can be determined by calculating enrichment factors, for example, using a computer-readable medium. The enrichment factor may be, or include, a comparison of the difference or ratio of the amount of products produced by selected and enriched cells compared to another cell, e.g., unselected or unenriched cells, wild-type cells, parental cells, etc. In some cases, the enrichment factor may include the ratio of the occurrence of variants (e.g., genetic variants) in a population of selected cells compared to an unselected population. The enrichment factor may serve as a surrogate for functional performance (e.g., functional performance). In parallel, sequence information can be obtained from enriched cells (e.g., those yielding the best functional performance) to determine gene sequences and / or genomic mutations. The identified gene sequences (or genomic mutations) can then be assigned to or paired with enrichment factors, thereby generating structure-function relationships between gene sequences and functional properties.
[0168] In some cases, the enrichment factor may be a comparison of the number of occurrences of a variant (e.g., a genetic variant) in an unselected sample compared to a selected sample. For example, the abundance of each genetic variant (e.g., identified using sequencing) may be determined in an unselected cell population as well as in a selected enriched cell population, and the enrichment factor may be calculated by taking the ratio or difference of the abundance of the genetic variant in enriched cells to the abundance of the genetic variant in unselected or unenriched cells. In such cases, an unselected sample or a portion thereof can be subjected to sequencing to identify genetic variants in the unselected sample, and the abundance of a particular genetic variant can be calculated, for example, by determining the number of occurrences of that particular genetic variant among all genetic variants in the unselected sample. Similarly, a selected sample or a portion thereof can be subjected to sequencing to identify genetic variants in the selected sample, and the abundance of a particular genetic variant can be calculated, for example, by determining the number of occurrences of that particular genetic variant among all genetic variants in the selected sample. The abundance of a particular genetic variant in an unselected sample compared to a selected sample may indicate the enrichment of that particular genetic variant from the enrichment process. For example, for a particular genetic variant X that appears 10 times out of 100,000 sequences in an unselected sample (or at a frequency of 1:10,000) and 1 time out of 100 sequences in a selected sample, the enrichment factor is (1 / 100) / (1 / 10,000) or 100 times. The enrichment factor may be useful as a surrogate or estimate of the change in functional performance (e.g., the amount of product produced or the activity level of a protein or enzyme). Alternatively, the enrichment factor may be calculated by the number of occurrences of the genetic variant in the selected sample compared to an unselected example. It will be understood that the enrichment factor may be calculated by comparing the number of occurrences of the variant in a selected or enriched sample compared to (i) an unselected sample or (ii) a sample obtained from a previous enrichment.
[0169] The enrichment factor may be approximately 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 100, 500, 1000, 10,000, 100,000, 100,000, 1,000,000, or more than 10,000,000. The enrichment factor may also be in the range of approximately 10 to 100, 100 to 1,000, etc.
[0170] In some examples, the molecular discovery process may not involve cells and may include collecting or extracting different discovered molecules, subjecting them to one or more application conditions, and isolating molecular variants with desired properties. Further characterization of the different discovered molecules (e.g., chemical or biochemical analysis) can be performed. For example, discovered molecules with desired properties (e.g., metabolic activity exhibiting desired physical or chemical properties) can be subjected to chemical or elemental analysis such as imaging (and image analysis), spectroscopy, mass spectrometry, gravimetric analysis, atomic spectroscopy, nuclear magnetic resonance (NMR), chromatography, plasma mass spectrometry, X-ray photoelectron spectroscopy (XPS), and Auger electron spectroscopy. The obtained data (e.g., chemical composition, physical properties, etc.) can be used to predict which structural elements (e.g., elements, chemical composition, etc.) possess the desired functional properties.
[0171] A molecular discovery unit can be used to discover new molecules (e.g., biomolecules such as proteins or peptides) by introducing a library of molecules (e.g., a genetic library containing multiple genetic variants or DNA sequences encoding different target proteins, or a molecular library containing multiple molecular variants encoding different proteins or resulting in functional properties) into cells or cell populations, culturing the cells under one or more application conditions as needed, and screening the cells for desired functional properties. Cells or cell populations exhibiting the desired functional properties (e.g., expression of a specific target protein under specified application conditions, or other phenotypes or metaboloids) can be isolated and further processed or analyzed (e.g., using sequencing). Iterative or recursive cycles of such processes can be useful for generating large amounts of data on the chemical or structural composition (e.g., protein motifs, DNA, or amino acid sequences) and function or performance (e.g., having desired properties, phenotypes, metaboloids, etc.) of molecules under a given set of application conditions. In some examples, additional inputs of biological diversity (e.g., via gene recombination and introduction of recombinant genes into cells, or via mutagenesis) can be introduced during recursive cycles. Alternatively or additionally, variety of conditions (e.g., changes in application or culture conditions) may be introduced during the recursive cycle.
[0172] In some cases, it may be useful to use a molecular discovery unit to screen various application conditions. In such cases, reaction vessels containing molecular mutations (which may be naturally occurring or induced, as described herein) as needed can be subjected to different application conditions. Non-limiting examples of different application conditions include temperature, growth or nutrient medium composition, reaction buffer composition, substrate concentration, oxygenation, salinity, pH, carbon source, buffer concentration, culture period, matrix composition, protein concentration, attachment or immobilization of molecules or variants to a support (e.g., beads or other solid supports), and aeration. For example, in molecular discovery in or using multiple cells, multiple cells can be cultured under a specified set of fermentation conditions before distributing the cells into the reaction vessel, or the cells can be cultured under specified fermentation conditions within the reaction vessel. The fermentation conditions may be the same or different across a subset or all of the reaction vessels. In another example, multiple cells can be subjected to application conditions that include treatments (e.g., drug or small molecule treatment, presence of substrates, various reaction conditions, etc.) that may affect the performance or desired properties of the multiple cells (e.g., target protein expression levels, enzyme activity levels). Varying the application conditions and observing the effects on the performance of multiple cells may be useful in understanding how environmental variations affect structure or function, and / or can be used to discover novel or optimized molecules that possess desired properties under given application conditions.
[0173] In the case of biomolecule discovery within or using cells, cultured cells can be screened in the reaction vessel of a molecular discovery unit, and cells possessing desired functional properties (e.g., expression of a target protein variant, enzyme activity, metabolism, or phenotypic trait) under specified application conditions can be isolated, collected, and analyzed (e.g., via sequencing) to identify genetic or protein variants (e.g., using nucleic acids or amino acid sequences). Genetic or protein variants possessing desired properties (e.g., expression of a target protein under specified application conditions, or other phenotypic or metabolic traits) can be isolated. Information on the genetic or protein variant (e.g., nucleic acid or amino acid sequence) and its corresponding performance (e.g., target protein expression level) can be generated. The process can be repeated in recursive cycles so that each cycle generates data on structural or sequence elements and their corresponding performance. In some examples, additional inputs of biological diversity (e.g., via gene recombination and introduction of recombinant genes into cells, or via mutagenesis) can be introduced during recursive cycles to generate additional data on the effects of such structural or sequence elements on structural or sequence elements and their performance. Alternatively or additionally, a variety of conditions (e.g., changes in the applicable conditions) may be introduced during the recursive cycle.
[0174] Different discovered molecules may be included in libraries (e.g., protein libraries, peptide libraries, biomolecular libraries), or different discovered molecules may be encoded in a genome variant library, which may include genetic variants from naturally occurring, generated, or synthesized sources. For example, a naturally occurring gene library may include DNA sequences collected from samples (e.g., biological samples, soil samples, etc.) and may constitute or include an environmental DNA (eDNA) library. In other examples, a library may include synthetically produced genetic variants. For example, genetic variants can be generated by mutagenesis (random mutagenesis (e.g., exposure to mutagens such as ultraviolet, X-ray, and gamma irradiation, chemical mutagens such as ethyl methyl sulfonate, reactive oxygen species, deamination agents, polycyclic aromatic hydrocarbons, alkylating agents (e.g., ethyl nitrosourea, nitrosamines), aromatic amines, alkaloids, bromine, sodium azide, psoralens, benzene, metals (e.g., arsenic, cadmium, chromium, nickel) or biological agents (e.g., transposons, viruses, bacteria)), saturated mutagenesis, cyclic substitution), multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), CRISPR-guided DNA polymerase, error-prone polymerase chain reaction (epPCR), or other genome engineering approaches. In some examples, a genome variant library is or includes target genome libraries such as whole-genome mutagenesis libraries, genome shuffle libraries, promoter swap libraries, and transposon libraries (e.g., promoter insertion libraries).In some examples, a target genome library contains or codes for one or more members selected from the group consisting of promoters, 5' untranslated regions, 3' untranslated regions, ribozymes, RNA stability sequences, secreted peptides, fusion proteins, DNA-binding domains, protein-protein-binding regions, codon-optimizing genes, codon-randomization genes, terminators, and non-homologous end joins. In some examples, a genome variant library may be a single-gene random mutation library, a site-saturated library (e.g., a protein or peptide library in which every amino acid is mutated at each site, with each of 20 amino acids being mutated), a small insert metagenomic library, a large insert metagenomic library, or a combination of these approaches. In some examples, the different discovery molecules are DNA-coding molecules. DNA-coding molecules may be included in libraries such as metagenomic libraries, DNA-coding libraries, libraries from one or more organisms, or libraries created by function generators. A DNA-coding molecule library may include randomly generated DNA-coding molecules, naturally occurring DNA-coding molecules, or a combination thereof.
[0175] Similar to the strain optimization unit, multiple experimental conditions can be screened simultaneously using the systems and methods described herein. For example, the molecular discovery unit can be configured to simultaneously screen or analyze millions of reaction volumes containing different discovery molecules or molecular variants subjected to the same or different application conditions. Beneficially, the molecular discovery unit can provide multiplexed multiparametric information regarding molecular variants and the environmental (or application) conditions of the molecular variants (also referred to herein as “discovered molecules”). Furthermore, the molecular discovery unit may be useful for discovering novel molecules with desired performance criteria (e.g., DNA-encoded proteins), such as threshold enzyme activity levels or other metabolic or phenotypic traits, or molecules with desired properties such as the production of target proteins under specific application conditions.
[0176] Reaction vessels: The reaction vessels of the molecular discovery unit may be, for example, droplets or plugs, wells (e.g., microwells, nanowells, picowells, etc.), vials, tubes, flasks, or other containers. In some examples, the reaction vessels are contained within a microfluidic device, which may contain or be part of a consumption unit. In some examples, the reaction vessels are members of an array, such as a microwell array or plate, or a nanowell array or plate. Multiple types of reaction vessels (e.g., droplets and wells) may be used simultaneously or sequentially for any of the processes described herein. The reaction vessels may be physically separate or distinct, or they may be able to merge or combine.
[0177] Each reaction vessel may contain different discovery molecules for screening. For example, as with strain optimization units, it may be useful to screen a library of genetic variants encoding different proteins for a desired functional property (e.g., biomolecular (e.g., enzyme) activity, biomolecular interactions, phenotype, or metabotropy), and therefore each reaction vessel may be configured to contain different genetic variants for screening. In some cases, each reaction vessel contains one or more cells having a genetic variant. For example, each reaction vessel may contain a single cell or multiple cells having a single genetic variant that is different from the genetic variants of other cells in other reaction vessels. Alternatively or additionally, each reaction vessel may contain a clonal population of cells having a single genetic variant. In other examples, it may be useful to screen different genetic variants within a single reaction vessel, and therefore the reaction vessel may contain a mixture of cells having different genetic variants. In such examples, each reaction vessel may contain combinations of cells and genetic variants.
[0178] As described above, reaction vessels may contain a single cell or multiple cells and can be configured to be evaluated in a high-throughput manner. For example, a molecular discovery unit can be configured to screen at least 100 cells / min, at least 500 cells / min, at least 1,000 cells / min, at least 5,000 cells / min, at least 10,000 cells / min, at least 50,000 cells / min, or at least 100,000 cells / min or more. Alternatively or additionally, reaction vessels can be screened in a high-throughput manner. For example, a molecular discovery unit can be configured to screen at least 10 reaction vessels / min, at least 100 reaction vessels / min, at least 500 reaction vessels / min, at least 1,000 reaction vessels / min, at least 5,000 reaction vessels / min, at least 10,000 reaction vessels / min, at least 50,000 reaction vessels / min, or at least 100,000 reaction vessels / min. In some examples, the molecular discovery unit can be configured to screen at least 10 reaction volumes / min, at least 100 reaction volumes / min, at least 500 reaction volumes / min, at least 1,000 reaction volumes / min, at least 5,000 reaction volumes / min, at least 10,000 reaction volumes / min, at least 50,000 reaction volumes / min, and at least 100,000 reaction volumes / min.
[0179] Any number of reaction vessels can be produced or used within a given period. In some examples, it may be beneficial to screen different numbers of reaction vessels (or volumes). Therefore, the number of reaction vessels (or volumes) used or produced can be adjusted according to the appropriate application or experimental objective. For example, if a large number of separate reaction volumes are useful for a screening purpose (e.g., for molecular discovery), then at least 10,000 reaction volumes, at least 100,000 reaction volumes, at least 1,000,000 reaction volumes, at least 10,000,000 reaction volumes, at least 100,000,000 reaction volumes, or more reaction volumes may be used. In other examples, a smaller number of separate reaction volumes may be useful or required (e.g., for targeted screening, or for the rational design of an experiment using a limited number of species for screening). In such cases, reaction volumes of up to 100,000,000, 10,000,000, 1,000,000, 100,000, 10,000, or less may be used or produced.
[0180] Similar to the strain optimization unit, the reaction vessel of the molecular discovery unit may be contained within a microfluidic device and may accommodate any useful volume. In the case of droplets, as described above, the droplets may contain any useful composition.
[0181] Assays: In some cases, the screening of biological variants may involve the use of one or more screening operations. Screening operations may include assays, e.g., direct assays, indirect assays, binding assays, reporter assays (e.g., assays based on reporter cells), fluorescence assays, chromogenic assays, or colorimetric assays. One or more immunological assays may be used, including but not limited to enzyme-linked immunosorbent assays (ELISA), sandwich ELISAs, and proximity ligation assays. In some cases, a reporter molecule or selection molecule is used for screening. The assay molecule or reporter molecule or selection molecule may be useful for detecting the presence of genetic variants or functional characteristics (e.g., phenotype or metabotropy) of the biological variant. For example, if the desired biological variant is a cell containing enzyme activity, the cells in the reaction vessel may be screened in the presence of a detection reagent, e.g., a chromogenic or fluorescent substrate that changes color or fluoresces when the substrate is converted to a product by the desired enzyme activity, or otherwise alters the surrounding environment. In such an example, a reaction vessel exhibiting a change in color or fluorescence may indicate that the cells possess the desired enzyme activity, and that the cells can subsequently be isolated and further analyzed (e.g., via sequencing, protein analysis, etc.) to determine information about the biological variant (e.g., DNA or RNA sequence, protein sequence, or composition). In another example, the desired biological variant may include a selection molecule or a reporter molecule, for example, in the case of a selection molecule, only cells with the desired traits can survive under given culture conditions (e.g., only cells with a transgene or recombinant gene containing antibiotic resistance can survive in a growth medium containing antibiotics). The surviving cells can be screened for the desired functional properties, and those with the desired functional properties can be subjected to further analysis. In yet another example, only cells with the desired traits may express a reporter molecule (e.g., a fluorescent or detectable protein) indicating the presence of the transgene or recombinant gene.In such cases, cells exhibiting desirable traits or functional characteristics can be isolated from the reaction vessel and subjected to further analysis, as described herein. In other cases, assays (e.g., immunoassays, binding assays, etc.) can be used to determine the presence of a target molecule (e.g., protein, nucleic acid molecule, lipid, carbohydrate, metabolite, or combination thereof) in each genetic variant, and the genetic variant containing the target molecule can be isolated and further analyzed. For example, an assay may involve the addition of one or more detection reagents, such as pH adjusters, ionic strength or ionic concentration adjusters, ions, substrates, inhibitors, enzymes, or competitors.
[0182] One or more screening operations may include imaging or detection of signals within the reaction vessel. For example, if the desired biological variant is a cell containing enzyme activity, the cells in the reaction vessel may be imaged in the presence of a detection reagent, such as a chromogenic or fluorescent substrate that changes color or emits fluorescence when the substrate is converted to a product by the desired enzyme activity, or otherwise alters the surrounding environment. The acquired images can be analyzed by a computer-readable medium, for example, to obtain signal intensity values. Based on the imaging results, components of the strain optimization unit may be configured to automatically sort cells or reaction vessels, for example, by moving cells or reaction vessels to different locations. Imaging may be performed at any useful or easy-to-use step, for example, before, during, or after distribution, culture (where applicable), screening, sorting, etc.
[0183] Further analysis can be performed on selected biological variants (e.g., selected cells exhibiting desired characteristics). Such analyses include, but are not limited to, protein analysis (e.g., mass spectrometry, protein assays), nucleic acid analysis (e.g., sequencing), lipid analysis (e.g., Raman scattering or spectroscopy), imaging (and image analysis), spectroscopy, chemical composition analysis (e.g., NMR, XPS, chromatography, etc.), or other analytical methods or screenings, as described herein. Such analytical methods may be useful in determining the identity of genetic variants by obtaining information, for example, on DNA sequences or genes, amino acid sequences, or protein properties (e.g., post-translational modifications). Such information can then be collected and stored in a database, or, if necessary, reintegrated with a molecular discovery library possessing any functional characteristics such as desired biological activity, phenotype, or metabolic traits.
[0184] In some examples, the molecular discovery unit may include or be coupled to a detector. The detector may be used to determine the presence or absence of a target molecule or functional property in a cell or reaction vessel, enabling screening and / or selection of desired biological variants. The detector may include one or more analytical instruments, such as a mass spectrometer, Raman spectrometer, near-infrared (NIR) spectrophotometer, Fourier transform infrared (FTIR) spectrometer, spectrophotometer (e.g., fluorescence or absorbance plate or cuvette reader), piezoelectric sensor, etc. The detector may be configured to count cells (e.g., cell counter, flow cytometer). The detector may include one or more optical components, and in some examples, it may include a microscope (e.g., stereoscopic mirror, compound microscope, fluorescence microscope, polarizing microscope, confocal microscope, differential interference microscope, super-resolution microscope, etc.), or components or combinations thereof. Detectors may be used to detect physical properties of cells or reaction vessels, including but not limited to magnetism, impedance, topography or geometric shape, scattering, morphology, absorbance, fluorescence, color, cell clustering, and cell or colony size. Detectors may be point detectors, array detectors, or image detectors. Detectors may be configured to collect point signals, arrays, point measurements, or images.
[0185] Moving Actuator: In some examples, a reaction vessel moving actuator is used to move the reaction vessel (e.g., toward a detector). The vessel moving actuator may be a robotic manipulator, flow generator, acoustic drop generator, optical tweezers, thermal drop on demand, piezoelectric drop on demand, lithography unit, digital microfluidic device, or a combination thereof. The moving actuator may comprise a stage and any necessary moving components such as screws, gears, threads, motors, pumps, compressors, resistors, transistors, springs, pulleys, etc. In some examples, the reaction vessel is positioned on or adjacent to a substrate connected to the moving actuator.
[0186] Molecular Library: A molecular discovery unit may include a discovery molecular library, which may be used in the processes described herein. The molecular discovery library may include multiple different genetic variants, as described herein. The multiple genetic variants may originate from a first strain. For example, cells of a first cell line (e.g., bacteria, yeast, animals, plants, etc.) can be subjected to conditions sufficient to produce multiple different genetic variants (e.g., via targeted or untargeted mutagenesis), such that each genetic variant is distinct. Each genetic variant of the multiple genetic variants may include, for example, DNA or RNA having a different sequence from other genetic variants of the multiple genetic variants. Multiple different genetic variants may encode mutations in the same biomolecule (e.g., DNA, RNA, peptides, proteins, etc.). For example, multiple different genetic variants may include different DNA sequences encoding mutations in RNA sequences that are later translated into proteins. Mutations in different DNA sequences in the library may result in different RNA sequences, and therefore translated proteins or peptides. In some examples, the protein or peptide encoded by a genetic variant (which may have mutations in DNA and / or RNA) is an enzyme. In such cases, the enzyme may form or be part of a metabolic pathway. Enzyme or metabolic activity can be used as a screening parameter or performance requirement during the molecular discovery process.
[0187] In some cases, the molecular discovery unit is configured to screen molecular variants and select and / or classify desired biological variants (e.g., molecular variants) based on functional characteristics (e.g., phenotypic or metabotropic traits or activity) or measurements. Phenotypic or metabotropic measurements may, in non-limiting examples, include protein or enzyme activity, protein expression levels, or the quantity or amount of by-products or metabolites. For example, a desired biological variant may have a specified or minimum threshold enzyme activity. In such cases, molecular variants in a molecular discovery library can be screened by the molecular discovery unit, and only molecular variants with the specified or minimum threshold enzyme activity are selected for further analysis. The molecular variants can be subjected to further analysis (e.g., sequencing, protein analysis, chemical analysis) to determine the nucleic acid or protein sequence or chemical composition of the enzyme of interest. In some examples, the sequence or composition of the molecular variant is added to a database and / or reintegrated into a molecular discovery library with functional characteristics or performance. Therefore, molecular discovery units can be useful in correlating sequence or structural information with functional properties or performance, or in elucidating non-standard structure-function relationships, for example, in discovering sequences with greater diversity or less homology that have similar functional performance under a given set of conditions.
[0188] A molecular discovery library may contain a number of molecular variants useful for screening or discovering molecules that have or produce desired functional properties. A molecular discovery library may contain at least 100 different discovered molecules, at least 1,000 different discovered molecules, at least 10,000 different discovered molecules, at least 100,000 different discovered molecules, at least 1,000,000 different discovered molecules, at least 10,000,000 different discovered molecules, at least 100,000,000 different discovered molecules, or more. As described herein, the molecular discovery process or integration system may be recursive or iterative. In such cases, the number of different discovered molecules screened in each iteration may vary; for example, the first cycle may screen 10,000,000 different discovered molecules, and the second cycle may screen 100,000 different discovered molecules. Conversely, the number of different discovered molecules can increase between cycles, for example, by introducing new biological diversity or variants (e.g., new strains, new and different discovered molecules, etc.) and / or conditional diversity (e.g., different fermentation conditions for some or all of the newly introduced or screened biological variants).
[0189] (iii) Molecular Optimization Unit In some examples, the systems and methods disclosed herein utilize molecular optimization units, which may be useful for screening and optimizing molecules (e.g., proteins, peptides, or other biomolecules) for functional properties. The systems and methods described herein may also utilize the use of one or more machine learning algorithms that input information or data into a library of variants and output a designed variant configured to perform the function in an improved manner.
[0190] Molecular optimization units can be useful not only for optimizing desired functional properties by manipulating the structure of a molecule, but also for screening molecular variants for desired functional properties and predicting molecular structure-function relationships (e.g., biomolecules or polymers, such as proteins, peptides, and nucleic acid sequences). In one example, a method for molecular optimization may include (a) obtaining structural performance (e.g., sequence information) and functional performance data of a molecule (e.g., protein, enzyme), and (b) generating predictions of the structure of an optimized molecule having functional performance above or below a threshold performance criterion using a computer-readable medium. In some examples, the method may further include (c) introducing the predicted structure into a library of variants (e.g., genetic variants, which may include recombinant or transgenes containing gene sequences related to or encoding target proteins of desired functional properties), and (d) screening the library of variants, which may be repetitive in some examples. In some examples, this method involves using a computer-readable medium (e.g., a model, simulator) to predict the structure of an optimized molecule (e.g., gene sequence or amino acid sequence, chemical composition, etc.) that has desired functional properties exceeding a performance threshold. In some examples, the computer-readable medium includes a machine learning algorithm.
[0191] In another example, a method for molecular optimization may include (a) providing a library of molecular or strain variants (e.g., a DNA-coding library encoding one or more peptides or proteins, a library of cells containing nucleic acid molecules (e.g., aptamers) with functional performance, or molecular (e.g., polymer or peptide) variants) and (b) screening the library of variants, which may be repetitive in some examples. The screening may include selecting variants that have desired functional properties or that meet functional performance criteria. For example, in the case of proteins or peptides, enzyme activity or binding properties can be monitored, and those with desired enzyme activity or binding properties can be selected and further optimized as needed.
[0192] In some examples, a molecular optimization unit includes a computer-based component (e.g., computer-readable media) that takes one or more datasets as input and outputs predictions of structure-function relationships for molecules or structures (e.g., sequences, motifs, chemical composition, etc.) that are predicted to yield desired functional properties. The datasets may include outputs from other integration units (e.g., molecular discovery units, strain optimization units). For example, data from a strain optimization unit, such as which fermentation conditions and genetic variants yield desired functional properties, can be used in a molecular optimization unit to generate predictions. Similarly, a molecular optimization unit can use data from a molecular discovery unit, such as sequence or structural information, material or biological origin, and application conditions of molecules with desired functional properties. Conversely, data from a molecular optimization unit can be used in a strain optimization unit or a molecular discovery unit for the respective strain optimization and molecular discovery processes.
[0193] In some cases, the datasets input to the molecular optimization unit include data from databases such as publicly available databases. For example, for protein or peptide optimization, these may include DDBJ, GeneCards, InterPro, NCBI, POWL, PIR, PRF, SYSTERs, 3Dee, ArchDB, ASTRAL, BioMagResBank, CDDB, EMDataBank, Enzyme Structures, fPOP, Jena Lib, PMP, Proteopedia, MolMovDB, PDB, PISA, PMP, SBKB, TOPSAN, COG, Pfam, PRINTS, iProClass, ProDom, PROSITE, SMART, AutoPSI, CATH, CDD, CE, CL, FSSP, HOMSTRAD, MODBASE, PartsList, PDBe, SCOP, VAST, AARSDB, ASD, ASPD, BRENDA, DAVID, Catalytic Site Atlas, EF-hand CaBP, EcoCyc, ENZYME, Gene Protein information can be used from databases such as Ontology, GPRCD, Homeobox, KEGG, MatrixDB, MEROPS, MetaCyc, P2CS, PREX, Protein Kinase Resource, Protein Ontology, RNase P, SuperCYP, Transport DB, CPLA, P3DB, Phospho3D, PHOSIDA, LOCATE, eSLDB, LOCATE, NPD, NURSA, ORGe, PSORTdb, SUBA, 3dID, UniProt, GenBank, RefSeq, SwissProt, Protein, MEDLINE, BioSystems, Entrez, SRS, SCOPPI, ReLiBase, Protein3dHome, MIPS, IBIS, DOMMINO, DOMINE, DIMA, DIP, and BioGRID.
[0194] In some cases, both experimental data and data from databases are used in the molecular optimization unit. Data from databases (e.g., sequence information, structure-function relationships, protein-protein interactions) can be used in conjunction with empirical data (e.g., from other integration units or the molecular optimization unit) to predict molecular variants expected to possess desired functional properties or a certain threshold level of functional performance. These predicted molecular variants can then be screened in the molecular optimization unit or other integration units (e.g., a strain optimization unit or a molecular discovery unit).
[0195] In one non-limiting example, data from a molecular discovery unit or strain optimization unit containing structural or sequence data, such as the DNA sequence encoding the enzyme of interest or the amino acid sequence of the enzyme in a desired biological variant, as well as the enzyme's performance (from one or more recursive cycles), and / or fermentation or application conditions, can be input to the molecular optimization unit (e.g., a computer-enabled component or computer-readable medium of the molecular discovery unit). Data from one or more protein or enzyme databases (e.g., UniProt, DDBJ, etc.) can also be added to the molecular optimization unit (e.g., a computer-enabled component or computer-readable medium). The computer-readable medium of the molecular optimization unit may include one or more algorithms that use the inputs to output models or predictions in which parameters (e.g., structure, sequence, application conditions, fermentation conditions, etc.) affect performance (e.g., enzyme activity). The computer-readable medium may also output predictions in which parameters (e.g., gene sequence or variant, fermentation conditions) are likely to result in performance at or above a threshold performance criterion (e.g., baseline activity level).
[0196] In some cases, predicted gene variants can be incorporated into a gene library, such as a plasmid library, via recombination strategies such as restriction digestion, sequencing-by-synthesis, or mutagenesis as described elsewhere in this specification (e.g., exposure to mutagens such as ultraviolet light or ethyl methyl sulfonate). The gene library can then be introduced into one or more cells, for example, by transfection, transformation, or transduction, thereby creating multiple cells containing the genetic variants. The multiple cells can be cultured under one or more applied or fermentation conditions, as needed. As described elsewhere in this specification, the multiple cells can be screened in a high-throughput manner for desired functional properties or performance (e.g., enzyme activity). Strains exhibiting the desired functional properties or performance can be isolated and subjected to further analysis (e.g., sequencing) to identify the genetic variants. This information or data (e.g., functional performance and gene sequences) can then be re-entered into a computer-readable medium (e.g., simulation or model) to further improve the predictive quality. Such a process, like the one presented in stock optimization, can be executed iteratively or recursively until the desired performance criteria are met.
[0197] A molecular optimization unit may include an experimental module, which may include a number of separate molecular optimization reaction vessels, each having a different target protein (or molecular variant) or application conditions. As described herein, a method for molecular optimization may include obtaining structural (e.g., sequence information) and functional performance data of a molecule (e.g., protein, enzyme); generating predictions about the structure of an optimized molecule having functional performance above or below a threshold performance criterion; introducing the predicted structure into a library of variants (e.g., genetic variants, which may include recombinant or transgenes containing gene sequences related to desired functional properties); and screening the library of variants, which may be repetitive in some examples. In some examples, a method for molecular optimization may include obtaining a library of variants (e.g., genetic variants created from randomly mutated genes or operons); and screening the library of variants for desired functional performance or properties. In some examples, one or more operations may be repetitive. Thus, a molecular optimization unit may include a system for processing a large number of reaction vessels.
[0198] For example, molecular optimization units, like strain optimization units, can be used in directed evolutionary techniques and manipulations of cells or cell populations to possess desired functional properties. In such examples, a library of cells containing one or more polymer variants or a group of cells with genetic diversity (which may occur naturally or be induced, for example, through mutagenesis or transformation of transgenes or recombinant genes, transduction, etc.) can be screened in the reaction vessel of the molecular optimization unit under multiple applicable conditions as needed, and cells with desired functional properties can be selected, isolated, collected, and / or analyzed (e.g., via sequencing). Functional properties (e.g., production of the molecule of interest, activity level of the molecule of interest) can be measured using imaging or assays as described herein, and can be performed before, during, or after screening. In some examples, a portion of the library of biological variants may not be selected, or may be selected to locations different from those of biological variants exhibiting the preferred functional properties. For example, a sample may contain a library of biological variants, with some of the sample being selected and others remaining unselected. Unselected (or selectively selected) biological variants can serve as a control population or group for comparing or measuring increases or enhancements in functional properties from molecular optimization processes. In some examples, selected cells may be directed to a different number of classes so that multiple cell populations are obtained. For example, gating may be performed so that cells exhibiting the highest functional performance are directed to the first group, cells exhibiting the second highest performance are directed to the second group, and so on.
[0199] The sorted cells can be subjected to further characterization or analysis. Analysis of biological variants may include identification of genetic variants (e.g., purification and sequencing to determine nucleic acids or gene sequences, if necessary). Similarly, unsorted or differently sorted cells can be subjected to further characterization, for example, to obtain sequence information or data. In some examples, sequence information may be paired with or correlated with functional properties. In one example, sequence information may be paired with functional properties based on the abundance of the sequence in sorted and unsorted cells (e.g., sequence count), for example, using a computer-readable medium. Alternatively or additionally, the abundance of each polymer variant can be obtained in both sorted and unsorted cells. From the sequence and abundance data, enrichment data (e.g., enrichment factors) can be obtained, for example, by comparing the abundance of each polymer variant in sorted cells with the abundance of each polymer variant in unsorted cells. Functional data, sequence data, and condensed data can be individually or collectively input into a computer-readable medium (e.g., a machine learning algorithm) to then design and output predicted variants that are expected to have improved functional characteristics.
[0200] In some examples, a library of variants may contain one or more reference markers or sequences that can be used for screening. For example, a library of samples or variants may be screened based on the function of polymer variants, with at least one of the polymer variants containing a reference sequence. The reference sequence can be used to calibrate measurements; for example, the reference sequence may contain known sequences and known functional activities, which can be used to calibrate or normalize the screening process to known functional activities. The reference sequence may contain one or more biological variants that may encode the same or different polymers. For example, two reference sequences may contain different DNA sequences that encode the same polymer (e.g., a peptide or a protein).
[0201] In some examples, biological variants possessing desired characteristics (e.g., a target gene, protein expression under specified fermentation conditions, or other phenotypic or metabolic traits) can be isolated and re-screened by recombining them with multiple recombinant genes, e.g., by transforming, transfecting, or transducing the recombinant genes into cells, culturing the cells as needed, and isolating cells with the desired functional characteristics and, if necessary, more stringent performance requirements. The process may be repeated in recursive cycles so that each cycle has more stringent performance requirements (e.g., higher or lower metabolic activity, enzyme activity, or higher or lower expression of phenotypic traits). The recursive cycles and the operations contained therein may be performed over as many cycles as necessary to satisfy the desired performance requirements. In some examples, additional inputs of biological diversity (e.g., via gene recombination and introduction of recombinant genes into cells, or via mutagenesis) can be introduced during recursive cycles. Alternatively or additionally, variability of conditions (e.g., changes in the conditions applied) can be introduced during recursive cycles.
[0202] In some cases, it may be useful to screen for various application conditions using conditional diversity screening, e.g., molecular optimization units. In such cases, reaction vessels or molecular variants can be subjected to different application conditions, and, if necessary, molecular mutations (which may be naturally occurring or induced, as described herein) can be subjected to different application conditions. Non-limiting examples of different application conditions include temperature, growth or nutrient medium composition, reaction buffer composition, substrate concentration, oxygenation, salinity, pH, carbon source, buffer concentration, culture period, supply schedule, and aeration. For example, in molecular optimization in or using multiple cells, the multiple cells can be cultured under a specified set of fermentation conditions before distributing the cells into the reaction vessel, or the cells can be cultured under specified fermentation conditions within the reaction vessel. The fermentation conditions may be the same or different across a subset or all of the reaction vessels. In another example, multiple reaction vessels (including cells as needed) can be subjected to application conditions that may affect the performance or desired properties of the multiple reaction vessels (e.g., target protein expression levels, enzyme activity levels), including treatments (e.g., drug or small molecule treatment, presence of substrates, various reaction conditions, etc.). Varying the application conditions and observing the effect on molecular performance can be useful in understanding how environmental variations affect structure or function, and / or can be further used to discover novel or optimized molecules (e.g., target proteins or other biomolecules) that have desired properties under given application conditions.
[0203] In some examples, a molecular optimization process may involve screening different molecules, subjecting them to one or more application conditions, and isolating molecular variants possessing desired properties, without involving cells. For example, the desired molecular variant may exhibit a specific protein-protein interaction or metabolic activity. A large number of molecules can be screened using the processes described herein, and desired molecular variants exhibiting functional properties can be isolated and subjected to further characterization. Further characterization of different discovered molecules (e.g., chemical or biochemical analysis) may include, in non-limiting examples, chemical or elemental analyses such as mass spectrometry, gravimetric analysis, atomic spectroscopy, nuclear magnetic resonance (NMR), chromatography, plasma mass spectrometry, X-ray photoelectron spectroscopy (XPS), and Auger electron spectroscopy. The obtained data (e.g., chemical composition, physical properties, etc.) can be used to predict (e.g., using a computer-readable medium) which structural elements (e.g., elements, chemical composition, etc.) may possess desired functional properties or improved functional performance (e.g., protein-protein interactions or metabolic activity).
[0204] Different molecular variants of a molecular optimization unit (e.g., target protein variants) may be included in a library (e.g., a protein library, a peptide library, a biomolecular library), or different discovered molecules may be encoded in a genome variant library, which in some examples includes genetic variants that are generated or synthesized. For example, in the case of generated or synthesized genetic variants, genetic variants may be generated by mutagenesis (random mutagenesis (e.g., exposure to mutagens such as ultraviolet, X-ray, and gamma irradiation, chemical mutagens such as ethyl methyl sulfonate, reactive oxygen species, deamination agents, polycyclic aromatic hydrocarbons, alkylating agents (e.g., ethyl nitrosourea, nitrosamines), aromatic amines, alkaloids, bromine, sodium azide, psoralens, benzene, metals (e.g., arsenic, cadmium, chromium, nickel) or biological agents (e.g., transposons, viruses, bacteria)), saturated mutagenesis, cyclic substitution), multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), CRISPR-guided DNA polymerase, error-prone polymerase chain reaction (epPCR), or other genome engineering approaches. In some examples, a genome variant library is or includes target genome libraries such as whole-genome mutagenesis libraries, genome shuffle libraries, promoter swap libraries, and transposon libraries (e.g., promoter insertion libraries). In some examples, a target genome library includes or codes for one or more members selected from the group consisting of promoters, 5' untranslated regions, 3' untranslated regions, ribozymes, RNA stability sequences, secreted peptides, fusion proteins, DNA-binding domains, protein-protein binding regions, codon-optimization genes, codon-randomization genes, terminators, and non-homologous end joins.In some examples, a genome variant library may be a single-gene random mutation library, a site-saturated library, a small-insert metagenomic library, or a large-insert metagenomic library. In some examples, the different discovered molecules are DNA-coding molecules. DNA-coding molecules may be included in libraries such as metagenomic libraries, DNA-coding libraries, libraries from one or more organisms, or libraries created by function generators. A DNA-coding molecule library may include randomly generated DNA-coding molecules, naturally occurring DNA-coding molecules, or combinations thereof.
[0205] Similar to the strain optimization unit, multiple experimental conditions can be screened simultaneously using the systems and methods described herein. For example, the molecular optimization unit can be configured to simultaneously screen or analyze millions of reaction volumes containing different molecular variants (e.g., target protein variants) subjected to the same or different application conditions. Beneficially, the molecular optimization unit can provide multiplexed multiparametric information about molecular variants and the environmental (or application) conditions of those molecular variants. Furthermore, the molecular optimization unit may be useful for screening molecules with desired performance criteria (e.g., DNA-encoded protein targets), such as threshold enzyme activity levels or other metabolic or phenotypic traits, or molecules with desired properties such as the production of target proteins under specific application conditions.
[0206] Reaction Vessels: The reaction vessels of the molecular optimization unit may be the same as or different from those of the strain optimization unit and the molecular discovery unit. The reaction vessels of the molecular discovery unit may be, for example, droplets or plugs, wells (e.g., microwells, nanowells, picowells, etc.), vials, tubes, flasks, or other containers. In some examples, the reaction vessels are contained within a microfluidic device, which contains or is part of a consumption unit. In some examples, the reaction vessels are members of an array, such as a microwell array or plate, or a nanowell array or plate. Multiple types of reaction vessels (e.g., droplets and wells) may be used simultaneously or sequentially for any of the processes described herein. The reaction vessels may be physically separate or distinct, or they may be able to merge or combine.
[0207] Each reaction vessel may contain different molecules (e.g., target protein variants) for screening. For example, as with strain optimization units, it may be useful to screen a library of genetic variants encoding different proteins for a desired functional property (e.g., biomolecular (e.g., enzyme) activity, biomolecular interactions, phenotype, or metabotropy), and therefore each reaction vessel may be configured to contain different genetic variants for screening. In some cases, each reaction vessel may contain one or more cells having a genetic variant. For example, each reaction vessel may contain a single cell or multiple cells having a single genetic variant that is different from the genetic variants of other cells in other reaction vessels. Alternatively or additionally, each reaction vessel may contain a clonal population of cells having a single genetic variant. In other examples, it may be useful to screen different genetic variants within a single reaction vessel, and therefore the reaction vessel may contain a mixture of cells having different genetic variants. In such examples, each reaction vessel may contain combinations of cells and genetic variants.
[0208] As described herein, reaction vessels may contain a single cell or multiple cells and may be configured to be evaluated in a high-throughput manner. Any number of reaction vessels can be generated or used within a given period. In some examples, it may be beneficial to screen different numbers of reaction vessels (or volumes). Thus, the number of reaction vessels (or volumes) used or generated can be adjusted according to the appropriate application or experimental objective. For example, if a large number of separate reaction volumes are useful for, for example, screening purposes (e.g., for molecular discovery), then at least 10,000 reaction volumes, at least 100,000 reaction volumes, at least 1,000,000 reaction volumes, at least 10,000,000 reaction volumes, at least 100,000,000 reaction volumes or more may be used. In other examples, a smaller number of separate reaction volumes may be useful or required (e.g., for targeted screening or for the rational design of experiments using a limited number of species for screening). In such cases, reaction volumes of up to 100,000,000, 10,000,000, 1,000,000, 100,000, 10,000, or less may be used or produced.
[0209] Similar to the strain optimization unit, the reaction vessel of the molecular discovery unit may be contained within a microfluidic device and may accommodate any useful volume. In the case of droplets, as described above, the droplets may contain any useful composition.
[0210] Assays: In some cases, the screening of biological variants may involve the use of one or more screening operations. Screening operations may include assays, e.g., direct assays, indirect assays, binding assays, reporter assays (which may be in vitro or in vivo), fluorescence assays, chromogenic assays, or colorimetric assays. One or more immunological assays may be used, including but not limited to enzyme-linked immunosorbent assays (ELISA), sandwich ELISAs, and proximity ligation assays. In some cases, a reporter molecule or selection molecule is used for screening. The assay molecule or reporter molecule or selection molecule may be useful for detecting the presence of genetic variants or functional characteristics (e.g., phenotype or metabotropy) of the biological variant. For example, if the desired biological variant is a cell containing enzyme activity, the cells in the reaction vessel may be screened in the presence of a chromogenic or fluorescent substrate that changes color or fluoresces when the substrate is converted to a product by the desired enzyme activity. In such examples, a reaction vessel exhibiting a change in color or fluorescence may indicate that cells possess the desired enzyme activity, and that these cells can subsequently be isolated and further analyzed (e.g., via sequencing, protein analysis, etc.) to determine information about the biological variant (e.g., DNA or RNA sequence, protein sequence, or composition). In other examples, the desired biological variant may include a selection molecule or a reporter molecule, for example, in the case of a selection molecule, only cells possessing the desired traits can survive under given culture conditions (e.g., only cells with a transgene or recombinant gene containing antibiotic resistance can survive in a growth medium containing antibiotics). The surviving cells can be screened for the desired functional properties, and those possessing the desired functional properties can be subjected to further analysis. In yet another example, only cells possessing the desired traits may express a reporter molecule (e.g., a fluorescent or detectable protein) indicating the presence of the transgene or recombinant gene. In such examples, cells exhibiting the desired traits or functional properties can be isolated from the reaction vessel and subjected to further analysis, as described herein.In other examples, assays (e.g., immunoassays, binding assays, etc.) can be used to determine the presence of a target molecule (e.g., protein, nucleic acid molecule, lipid, carbohydrate, metabolite, or combination thereof) in each genetic variant, allowing for the isolation and further analysis of the genetic variant containing the target molecule.
[0211] Further analysis can be performed on selected biological variants (e.g., selected cells exhibiting desired characteristics). Such analyses include, but are not limited to, protein analysis (e.g., mass spectrometry, protein assays), nucleic acid analysis (e.g., sequencing), lipid analysis (e.g., Raman scattering or spectroscopy), chemical composition analysis (e.g., NMR, XPS, chromatography, etc.), imaging (and image analysis), spectroscopy, or other analytical methods or screenings, as described herein. Such analytical methods may be useful in determining the identity of genetic variants by obtaining information, for example, on DNA sequences or genes, amino acid sequences, or protein properties (e.g., post-translational modifications). Such information can then be collected and stored in a database, or, if necessary, reintegrated with a molecular discovery library possessing any functional characteristics such as desired biological activity, phenotype, or metabolic traits.
[0212] In some examples, the molecular optimization unit may include or be coupled to a detector. The detector may be used to determine the presence or absence of a target molecule or functional property in a cell or reaction vessel, enabling screening and / or selection of a desired biological variant. The detector may include one or more analytical instruments, such as a mass spectrometer, Raman spectrometer, near-infrared (NIR) spectrophotometer, Fourier transform infrared (FTIR) spectrometer, spectrophotometer (e.g., fluorescence or absorbance plate or cuvette reader), piezoelectric sensor, etc. The detector may be configured to count cells (e.g., cell counter, flow cytometer). The detector may include one or more optical components, and in some examples, it may include a microscope (e.g., stereoscopic mirror, compound microscope, fluorescence microscope, polarizing microscope, confocal microscope, differential interference microscope, super-resolution microscope, etc.), or components or combinations thereof. Detectors may be used to detect physical properties of cells or reaction vessels, including but not limited to magnetic properties, impedance, topography or geometric shape, scattering, morphology, absorbance, fluorescence, color, cell clustering, and cell or colony size. Detectors may be point detectors, arrays, or image detectors.
[0213] Moving Actuator: In some examples, a reaction vessel moving actuator is used to move the reaction vessel (e.g., toward a detector). The vessel moving actuator may be a robotic manipulator, flow generator, acoustic drop generator, optical tweezers, thermal drop on demand, piezoelectric drop on demand, lithography unit, digital microfluidic device, or a combination thereof. The moving actuator may comprise a stage and any necessary moving components such as screws, gears, threads, motors, pumps, compressors, resistors, transistors, springs, pulleys, etc. In some examples, the reaction vessel is positioned on or adjacent to a substrate connected to the moving actuator.
[0214] Library: A molecular optimization unit may include a molecular variant (e.g., target protein variant) library, which may be used in the processes described herein. The molecular optimization library may include multiple different genetic variants or protein variants, as described herein. For example, the library may include multiple target protein variants encoded by different genetic variants. The multiple genetic variants may originate from a first strain. For example, cells of a first cell line (e.g., bacteria, yeast, animals, plants, etc.) can be subjected to conditions sufficient to produce multiple different genetic variants, such that each genetic variant is different. Each genetic variant of the multiple genetic variants may include, for example, DNA or RNA having a different sequence from other genetic variants of the multiple genetic variants. Multiple different genetic variants may encode mutations in the same biomolecule (e.g., DNA, RNA, peptides, proteins, etc.). For example, multiple different genetic variants may include different DNA sequences encoding mutations in RNA sequences that are later translated into proteins. Mutations in different DNA sequences in the library can result in different RNA sequences, and therefore proteins. In some cases, proteins encoded by genetic variants (which may have mutations in DNA and / or RNA) are enzymes. In such cases, the enzyme may form or be part of a metabolic pathway. Enzyme or metabolic activity can be used as a screening parameter or performance requirement during molecular optimization processes.
[0215] In some cases, the molecular optimization unit is configured to screen molecular variants and select and / or classify desired biological variants (e.g., molecular variants) based on functional properties (e.g., phenotypic or metabotropic traits or activity) or measurements. Phenotypic or metabotropic measurements may, in non-limiting examples, include protein or enzyme activity, protein expression levels, and quantities of by-products or metabolites. For example, a desired biological variant may have a specified or minimum threshold enzyme activity. In such cases, molecular variants in a molecular library (e.g., a target protein variant library) can be screened by the molecular discovery unit, and only molecular variants with the specified or minimum threshold enzyme activity are selected for further analysis. The molecular variants can be subjected to further analysis (e.g., sequencing, protein analysis, chemical analysis) to determine the nucleic acid or protein sequence or chemical composition of the enzyme of interest. In some examples, the sequence or composition of the molecular variants is added to a database and / or reintegrated into a molecular optimization library with functional properties or performance. In some examples, the sequence or composition of a molecular variant is used in a computer-readable medium to generate predictions about the performance of a molecule (e.g., a target protein variant) based on a given set of conditions (e.g., application conditions) and structure (e.g., sequence, chemical composition, etc.).
[0216] A molecular optimization library may contain any number of molecular variants useful for screening molecules that have or influence desired functional properties (e.g., via protein-protein interactions or signaling pathways), such as target proteins. A molecular optimization library may contain at least 100 different molecules, at least 1,000 different molecules, at least 10,000 different molecules, at least 100,000 different molecules, at least 1,000,000 different molecules, at least 10,000,000 different molecules, at least 100,000,000 different molecules, or more. As described herein, the molecular optimization process may be recursive or iterative. In such cases, the number of different discovery molecules screened in each iteration may vary; for example, the first cycle may screen 10,000,000 different molecules, and the second cycle may screen 100,000 different molecules. Conversely, the number of different discovered molecules can increase between cycles, for example, by introducing new biological diversity or variants (e.g., new strains, new and different discovered molecules, etc.) and / or conditional diversity (e.g., various fermentation conditions or application conditions for some or all of the newly introduced or screened biological variants).
[0217] Target Protein: Molecular variants may include genetic variants encoding the target protein. The target protein may be any useful protein or peptide, such as structural proteins, enzymes, surface receptor proteins, peptide hormones, therapeutic peptides, immune system components (e.g., antibodies, antibody fragments, antibody conjugates), bioactive peptides (e.g., lantric peptides, lasso peptides, non-ribosomal synthesized peptides), or other useful proteins or peptides. The target protein may also be industrial enzymes such as polymer-degrading enzymes (e.g., polyethylene terephthalate hydrolase), lipozymes, paratases, cellulases, amylases, resinases, amidases, and isomerases. The target protein may also be therapeutic enzymes such as lipases, DNAse, L-asparaginase, L-glutaminase, urokinase, uricase, collagenase, glucosidase, galactosidase, and tyrosinase. The target protein may also be a therapeutic protein such as an antibody, antibody fragment, or other binding protein. The target proteins are amylin, adiponectin, adrenocorticotropic hormone, angiotensin, antidiuretic hormone, natriuretic peptide, calcitonin, cholecystokinin, corticotropin-releasing hormone, cortistatin, enkephalin, endothelin, erythropoietin, follicle-stimulating hormone, galanin, gastric inhibitory polypeptide, gastrin, ghrelin, glucagon, releasing hormone, hepcidin, human chorionic gonadotropin, human placental lactogen, growth hormone, inhibin, insulin Phosphorus, leptin, lipotropin, luteinizing hormone, stimulating hormone, motilin, orexin, osteocalcin, oxytocin, pancreatic polypeptide, parathyroid hormone, pituitary adenylyl cyclase-activating peptide, prolactin, relaxin, renin, secretin, somatostatin, thrombopoietin, thyroid-stimulating hormone, thyroid-stimulating hormone-releasing hormone, vasoactive intestinal peptide, guanylin, uroguanilin, or other peptide hormones may also be peptide hormones.The target protein may be a structural protein such as actin, actinin, aggrecan, biglycan, cadherin, clathrin, collagen, decorin, elasticity, fibrin, fibrinogen, fibronectin, heparin, keratin, laminin, mucin, myelin, myosin, spectrin, tropomyosin, troponin, tubulin, vimentin, or vitronectin. The target protein may also be a protein or enzyme involved in DNA replication, transcription, or translation, such as polymerase (e.g., RNA polymerase), ligase, helicase, ribozyme, ribosome, spliceosome protein, or chaperone protein. The target protein may also be a signaling factor or cytokine (e.g., interleukin, interferon, tumor necrosis factor, transforming growth factor, immunomodulator, etc.). The target protein may also be a food-based or nutrient-based protein, such as ovalbumin or hemoglobin.
[0218] Datasets: One or more components of a molecular optimization unit may include computer-readable media, as described in more detail below. In some examples, the computer-readable media is a computer model configured to take data as input and generate predictions about the functional performance of structural aspects (e.g., sequences, peptide or protein motifs, chemical composition, etc.). The computer-readable media may be configured to take multiple datasets as input, which may come from a database or experimental module, for example, from an integration unit. For example, the computer-readable media may include a machine learning algorithm that takes functional data, sequence data, and / or enriched data (e.g., polymer variants) as input and outputs a design molecule that is expected to have improved functionality. The computer-readable media may also be configured to perform a variety of other functions, such as storing datasets, setting thresholds for the datasets to be stored, and communicating with one or more integration units for automated processing.
[0219] In some examples, when machine learning algorithms are used, the machine learning algorithms can include supervised learning algorithms, unsupervised learning algorithms, or reinforcement learning algorithms. Algorithms may include linear regression, logistic regression, decision trees, support vector machines (SVMs), naive Bayes, k-nearest neighbors, k-means algorithms, random forests, dimensionality reduction algorithms, gradient boosting algorithms, XGBoost, light gradient boosting algorithms, Catboost, etc., or combinations thereof. Machine learning algorithms can include natural language models and one or more neural networks. Natural language models can have one or more techniques applied for computer implementation, such as parsing or semantic analysis. For example, parsing can be used and may include subtasks such as tokenization, part-of-speech tagging, headword recognition and stemming, and stop word removal. Semantic analysis may include disambiguation of word meanings, relation extraction, etc.
[0220] In some examples, the experimental system of a molecular optimization unit may be integrated with other integration units. For instance, a molecular optimization unit may include a computer-readable medium that generates predictions about optimal molecules or genetic variants, and these optimal molecules or genetic variants may be introduced into a strain optimization library within a strain optimization unit. Thus, different functions can be performed by interacting integration units, as described below.
[0221] (iv) Integrated System Computer-Readable Media: In some examples, one or more integration units are integrated or combined via computer-readable media. In non-limiting examples, computer-readable media may perform one or more functions, including storing datasets from any of the integration units in a data repository, setting threshold parameters for data delivered to the data repository, and instructing any of the units to repeat input diversity screening. For example, data such as sequence information and functional performance from a strain optimization unit can be integrated into computer-readable media, which can then be used to generate molecular optimization predictions (e.g., providing genetic variants predicted to perform above a threshold criterion). In another example, data from a molecular discovery unit can be integrated into computer-readable media, which can then be used to generate predictions about molecular variants predicted to improve functional performance under given fermentation or application conditions.
[0222] In some cases, computer-readable media include quality control functions. For example, computer-readable media can be configured to store data that exceeds a quality threshold (e.g., in a database or data repository). Thus, computer-readable media can set a threshold parameter for data and apply that threshold to input data. Data that adheres to the threshold (e.g., exceeds the threshold) is delivered to the data repository. Non-limiting examples of threshold parameters include signal-to-noise ratio, statistical significance, target performance criteria such as enzyme activity levels, protein or molecular interactions, or the presence of phenotypes or metabolic traits.
[0223] System Automation: In some examples, a computer-readable medium is configured to automate one or more processes in one or more units of an integrated unit. In such cases, one or more integrated units may have a communication interface (e.g., wireless or wired communication interface) that communicates with the communication interface of the computer-readable medium. The computer-readable medium can issue commands or instructions to the integrated unit (or its components) (e.g., via the communication interface). For example, the computer-readable medium may be configured to control one or more reaction motion actuators (e.g., fluid handling units, flow generators, etc.) of an integrated unit by relaying one or more commands or instructions, thereby controlling the movement of the reaction vessel. Similarly, the computer-readable medium may be configured to control the supply of reagents (e.g., culture medium, reaction reagents, or detection reagents) to the reaction vessel using, for example, fluid handling units, flow generators, movement actuators, etc. The reaction motion actuators, computer-readable medium, and detectors may all be integrated and, in some examples, connected via a feedback loop.
[0224] For example, a feedback loop can be used to control the volume in each reaction vessel (e.g., reagent volume), which can be beneficial in ensuring consistent cell growth within each vessel. The feedback loop can be used to monitor the conditions in each reaction vessel and automatically adjust them (e.g., via a feedback controller) based on a set of predetermined conditions, such as temperature, pH, and culture medium components. Similarly, the feedback loop may include detection or image recognition software, which can be used to determine whether a given reaction vessel contains the desired number of cells, the correct volume, flow rate, etc. If droplets are used, the feedback loop may include a flow rate regulator to maintain the frequency of droplet formation. Reaction vessels that do not contain the desired number of cells may be excluded from further screening or analysis. Thus, a computer-readable medium can make a decision based on such feedback as to whether the cells or reaction vessels should be further screened or analyzed.
[0225] In some examples, a feedback loop can be used in the screening process described herein. The reaction vessel movement actuator can be configured to sort and isolate desired biological variants (e.g., sorting droplets in microfluidic devices using fluorescence or colorimetric activated cell sorting, magnetic sorting, dielectrophoresis, etc.). In such examples, a detector can collect a signal or image from each reaction vessel, and the signal or image may be processed (e.g., via a computer-readable medium), and reaction vessels exhibiting functional performance (e.g., fluorescence or colorimetric signals exceeding a threshold) can be automatically gated and / or sorted (e.g., to another chamber, vessel, or part of the apparatus). Thus, the reaction motion actuator (or sorter), detector (detecting the signal), and computer-readable medium can be integrated. In some examples, a quantity or proportion of reaction vessels and / or cells, e.g., a fixed proportion (e.g., the top of a user-defined percentage), can be sorted. In some examples, only the top percentage of cells or reaction vessels (e.g., the highest functional performance) can be collected and sorted for further analysis. For example, only the top 50%, 40%, 30%, 30%, 25%, 15%, 10%, 5%, 1%, 0.5%, 0.1%, 0.05%, 0.025%, 0.01%, etc. of cells or reactors can be selected for collection. In some examples, the reaction motion actuator can be configured to pause, delay, or otherwise modify the motion of the reaction vessel based on the output of a computer-readable medium using a feedback loop. For example, if an error is detected by a detector (e.g., an undetectable signal, suspected interference, etc.), the computer-readable medium can automatically detect the error and instruct the motion actuator to pause, delay, or modify the movement of the reaction vessel.
[0226] Similarly, a computer-readable medium can instruct one or more system components (e.g., a fluid handling unit, a flow generator, etc.) to automatically introduce additional reagents (e.g., detection reagents, or other reaction reagents, e.g., pH adjusters, ionic strength adjusters, ions, substrates, inhibitors, enzymes, competitors, oxygen, carbon dioxide, etc.) into the reaction vessel. Such integration of multiple system components can enable precisely defined reaction times, along with detection, analysis, sorting, etc., within a specified time frame. For example, the addition of a detection reagent (e.g., a substrate for an enzyme assay to measure enzyme activity levels) can be instructed by the computer-readable medium so that each reaction vessel is incubated with a certain amount of detection reagent for the same period before detection. Since incubation time can affect readouts (e.g., longer incubation times increase signal intensity as more substrate is converted to the product), precisely controlling the amount of detection reagent added and / or incubation time can prevent, for example, erroneous readouts from reaction vessels that are incubated for longer periods and do not necessarily exhibit higher functional performance.
[0227] Therefore, the methods and systems described herein can enable precise system control with minimal variation across the reaction vessels. For example, a computer-readable medium, when executed by one or more processors, can include instructions to the system to distribute a population of cells into multiple reaction vessels or reaction volumes to produce distributed cells, to assay the phenotypic or functional performance of the distributed cells, and to generate a signal about the distributed cells. The generated signal may have a low coefficient of variation across the reaction vessels. For example, for a given biological variant (e.g., a strain containing the same DNA sequence, a homogeneous cell population, etc.), the coefficient of variation (e.g., which can represent the quantity or activity level of a product, as measured by signal output) may be less than approximately 500%, less than approximately 400%, less than approximately 300%, less than approximately 200%, less than approximately 100%, less than approximately 90%, less than approximately 80%, less than approximately 70%, less than approximately 60%, less than approximately 50%, less than approximately 40%, less than approximately 30%, less than approximately 20%, less than approximately 10%, less than approximately 9%, less than approximately 8%, less than approximately 7%, less than approximately 6%, less than approximately 5%, less than approximately 4%, less than approximately 3%, less than approximately 2%, or less than approximately 1% or less. In some examples, the coefficient of variation may be in the range of values, for example, 1% to approximately 10%, or approximately 2% to approximately 8%, or approximately 10% to approximately 100%. Such coefficients of variation can account for different noise characteristics within the system, such as biological variability (e.g., naturally occurring performance differences in biology across the same strain), reagent addition and incubation times, and optical / detection noise. In some examples, the strain optimization, molecular optimization, or molecular discovery units described herein may have coefficients of variation of less than approximately 10%, demonstrating that the systems and methods used herein may be advantageous for accurately characterizing libraries of variants with low technical noise or variability.
[0228] In some examples, computer-readable media may include proportional-integral-derivative (PID) algorithms that can compensate for drift within the system. For example, PID algorithms can compensate for drift caused by deviations in mechanical equipment, liquid viscosity, or temperature. Drift compensation can be beneficial in preventing erroneous sorting decisions by the system caused by drift. Such compensation may include, for example, modifying the threshold gate used to sort reaction vessels or cells, or adjusting the target percentage of sorted cells (e.g., the top percentage of cells with the highest function). In some examples, computer-readable media may include proportional or proportional-integral controllers.
[0229] In some examples, computer-readable media can be configured to execute one or more commands related to the execution of a strain optimization process, a molecular discovery process, or a molecular optimization process. Computer-readable media can instruct each unit individually or collectively to perform one or more of the operations described herein. For example, a computer-readable unit can instruct a strain optimization unit to perform a screening of biological conditions under given experimental conditions (e.g., using a set of biological or genetic variants, application or fermentation conditions, etc.), to rerun the experimental conditions, or to repeat a diversity screening (e.g., rerunning an experiment with added genetic variants, application or fermentation conditions, or a combination thereof).
[0230] In some examples, a computer-readable medium can be configured to take one or more target parameters as input. These target parameters may include, for example, performance thresholds or desired functional performance such as enzyme activity, molecular or protein interactions, or binding affinity. In such examples, the computer-readable medium can be configured to execute one or more commands related to the execution of a strain optimization process, a molecular discovery process, or a molecular optimization process until the target parameter is reached. In one non-limiting example, a target parameter of a desired level of enzyme activity can be input to the strain optimization unit. The computer-readable medium can then be configured to iteratively and / or automatically execute the strain optimization process, inputting additional biological or conditional diversity as needed (e.g., by introducing new specific or non-specific genetic variants or fermentation conditions) and screening enzyme activity (e.g., using fluorescence assays) until the target threshold for enzyme activity is met. For example, the computer-readable medium can be configured to automatically introduce one or more point-based or specific mutations into cells at known loci(s), for example, using known or empirically derived structural (e.g., sequence)-functional relationships. In another example, a computer-readable medium can be configured to automatically introduce random mutations into cells at either a specified or unspecified locus. Further screening can be performed to isolate strains that meet a target threshold, and then analyze those strains or proteins.
[0231] Target Performance Parameters: One or more target performance parameters can be input into the system (e.g., one or more integrated units of an integrated unit) to screen for biological or molecular variants. The target performance parameters may relate to desired functional characteristics of the biological or molecular variant. For example, the desired functional properties may be the yield of the desired molecule (e.g., protein or peptide), the growth or proliferation rate of cells, the sensitivity of cells to conditions (e.g., pH, temperature, ionic concentration such as divalent cations such as calcium and magnesium, or fermentation conditions), cell viability, titer, molecular interactions (e.g., protein-protein interactions), stability (e.g., thermal stability, resistance to protease degradation, etc.), binding properties (e.g., affinity or binding activity to another molecule, substrate specificity), metabolic or phenotypic traits (e.g., enzyme activity, catalytic performance), protein expression level, quantity of by-products or metabolites, presence or degree of target protein folding, target protein to by-product ratio, modifications of the target protein (e.g., glycosylation, phosphorylation, SUMOylation, acylation, ubiquitination, acetylation, lipidation, methylation, hydroxylation, or other post-translational modifications), or other functional properties. In some examples, the target performance parameter may be selected as a condition parameter. For example, the desired biological variant may be cells in which the yield of the target protein is increased under specified fermentation conditions (e.g., hypoxia, low temperature, changes in ion concentration, etc.). In another example, the desired biological variant may be cells that grow under specified fermentation conditions.
[0232] Integration of data from integration units: In some examples, data from each of the integration units can be added to a computer-readable medium (e.g., within a data repository). For example, as described herein, one or more integration units may include or interact with a computer-readable medium. The computer-readable medium may be configured to take one or more datasets as input and output predictions such as molecular structure-function relationships and optimal operating conditions (e.g., application or fermentation conditions) for desired functional properties. The datasets input to the computer-readable medium may come from any one of the integration units. For example, a dataset may be generated from one of the screenings of biological or molecular variants from any one of the integration units (e.g., a strain optimization unit) and may include information about the desired biological variant and its properties (e.g., structure, sequence, composition, etc.) as well as conditional information (e.g., culture, fermentation, application, or reaction conditions). In some examples, datasets not generated from integration units may be included in or added to the computer-readable medium. For example, a dataset from a database can be input to the computer-readable medium.
[0233] Biological diversity and conditional diversity screening: In some embodiments, the integrated system may be input to a database, dataset, or data stream containing information on biological diversity screening and / or conditional diversity screening. Biological diversity data may include information on biological variants (e.g., genetic variants, molecular variants, etc., having different genetic elements) for input to biological diversity screening (e.g., screening to identify desirable biological variants), and the input may include, in non-limiting examples, strain information, DNA sequences, RNA sequences, amino acid sequences, protein modifications (e.g., post-translational modifications), functional properties (e.g., enzyme activity, yield, titer, growth rate, protein-protein interactions, binding properties), or combinations thereof. For example, a biological diversity data stream may include data generated from screening from a strain optimization unit, which can generate data on different gene (or genomic) variants that produce a desired target protein or have desired functional properties. Such data may include (i) the identity of each strain's gene (or genome) variant, e.g., the DNA or plasmid sequence and / or peptide or protein produced from each gene (or genome) variant, and (ii) the performance characteristics (e.g., enzyme activity, growth rate) of each genetic variant or the protein or peptide derived therefrom. Other biodiversity data from public sources or databases (e.g., protein structure and sequence, protein performance) may also be included in the biodiversity dataset.
[0234] Conditional diversity data may include data on one or more conditions used to generate a desired biological variant. For example, during strain optimization, one or more fermentation conditions (e.g., pH, nutrient levels, carbon dioxide levels, temperature, etc.) can be modified to achieve a desired functional performance in a set of genetic variants. These fermentation conditions, and in some cases functional performance, can constitute one or more conditional diversity datasets. Similarly, application conditions used in molecular optimization (e.g., for the production or optimization of a target protein) or molecular discovery may be included in conditional diversity data. Other conditional diversity data (e.g., application or fermentation conditions) from public sources or databases may also be included within conditional diversity datasets.
[0235] In some cases, biodiversity data and conditional diversity data are independently programmable datasets. For example, biodiversity data and conditional diversity data can be generated from a single experiment (e.g., strain optimization screening), but the generated data can be discretized into the experimental input parameters or variables. For example, the experimental input parameters or variables (e.g., genetic variants and fermentation conditions) may yield a set of data including the functional performance of each combination of genetic variants and fermentation conditions. However, a programmable dataset can be configured such that each parameter or variable is independently manipulable. In one example, a computer-readable medium can generate a parameterized (e.g., multiparametric) model that can analyze the respective relevant weights of variables (e.g., fermentation conditions and genetic variants) on strain performance. In some cases, biodiversity data or conditional diversity data includes both experimental data and data from databases (e.g., publicly available datasets).
[0236] Overall, using data from biological diversity screening and conditional diversity screening, desirable biological variants (e.g., genetic variants) can be identified based on target parameters (e.g., target performance parameters) input to the system.
[0237] Machine Learning Algorithms: In some examples, computer-readable media includes machine learning algorithms. A machine learning algorithm may comprise one or more neural networks that can be used to iteratively or sequentially integrate and analyze data generated from integration units or external sources (e.g., publicly available databases) to discover new features (e.g., genetic variants encoding proteins) or molecules that are predicted to have functional properties or target performance. A machine learning algorithm may be trained using datasets from large libraries, such as publicly available or publicly provided databases of known structure-function relationships of molecules that have been shown to exhibit functional properties, e.g., DNA or amino acid sequences. Alternatively or additionally, a machine learning algorithm may be trained using datasets generated from one or more integration units.
[0238] A machine learning algorithm can include any number or type of useful algorithms. For example, a machine learning algorithm can include Elastic Net Regression Generalized Linear Model (GLMNET), Support Vector Machine Regression (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Multilayer Perceptron (MLP), Convolutional Neural Network (CNN), Deep Neural Network (DNN), or a combination thereof. A machine learning algorithm may include supervised learning, unsupervised learning, support vector machines, or other algorithms. A machine learning algorithm can be used to predict or generate biological variants that can achieve a desired functional characteristic. For example, a machine learning algorithm can be configured to output at least 1, at least 10, at least 100, at least 1,000, at least 10,000, at least 100,000, at least 1,000,000 or more biological variants that can have the functional characteristics of a predicted or threshold quantity.
[0239] In one example, a machine learning algorithm can be used to predict biological variants (e.g., genetic variants) or their structural components (e.g., molecular variants such as gene sequences, amino acid sequences, or motifs) that have desired or improved functional performance (e.g., enzyme activity) that has not yet been empirically tested or determined. The predicted biological or molecular variants may be novel molecules or sequences, known molecules or sequences, or combinations thereof. The machine learning algorithm can generate such predictions by inputting data from each of the integration units as well as publicly available datasets. In some examples, the input data may be processed, e.g., cleaned, transformed, normalized, standardized, etc. The predicted variants can then be empirically tested for performance in one of the integration units. Functional performance is recorded and reintegrated into the machine learning algorithm to further refine the prediction algorithm. For example, the prediction algorithm can be iteratively tuned or trained using data that supports the correctness of the prediction and / or data that proves the error of the prediction. This process can be repeated many times, for example, over any number of generations, until machine learning performance is obtained, for example, until the machine learning algorithm outputs the predicted molecule with a specified accuracy or performance criterion.
[0240] For example, training data may include data from a strain or molecular optimization unit. For instance, a library of variants, such as cells containing genetic variants encoding polymers (e.g., proteins), can be generated using saturated mutagenesis, for example, in which each position of the polymer, such as a peptide or protein, can be mutated to all 20 amino acids. Further random mutagenesis and recombination may be performed as needed. The library of variants can be screened using the methods provided herein, and a machine learning algorithm can be trained by inputting data (e.g., sequence data, enrichment data, functional performance data (e.g., for a given nucleic acid sequence encoding a polymer, e.g., protein or peptide, or for a protein or peptide sequence, the quantity or amount of products produced, enzyme activity, etc.), screening or assay conditions, etc.) to predict new variants that are expected to have improved functional properties.
[0241] In some examples, a machine learning algorithm integrates data from each of the integration units. For instance, a molecular discovery unit might take one or more molecules and application conditions as input, screen the molecules for desired functional activity, and output molecular identity (e.g., DNA or amino acid sequence) and the functional performance of each molecule. These outputs can then be provided as input to a molecular optimization unit. The molecular optimization unit can then use, if necessary, biological diversity and / or conditional data from a database (e.g., a DNA or protein database such as UniProt) to generate predictions that biological variants (e.g., genetic variants of strains encoding a target protein) or molecular variants (e.g., proteins, motifs, structures, compositions) are expected to have a threshold level of performance.
[0242] In some examples, the output prediction may include additional variants that have sequence homology with variants expected to have threshold-level performance. For example, the prediction may include variants with at least 70%, at least 80%, 90%, or more sequence homology with respect to the primary predicted variant. Homologousity can be determined, for example, by aligning the sequences of two or more nucleic acid molecules or proteins and determining the degree of overlap. Additional variants can also be screened (e.g., using a strain optimization unit) to obtain additional data on the performance of such variants. The additional data can be used to generate an additional library of sequences or parts of sequences (subsequences) containing information about key residues that are considered important to the functional performance of a molecule (e.g., a target protein). In some embodiments, the subsequence library can be generated by manipulating an existing library of genes or sequences encoding key residues (e.g., via gene shuffling or other recombination strategies).
[0243] Biological or molecular variants may be incorporated into a library (e.g., a gene library in the case of biological genetic variants). In some examples, the library is introduced into multiple cells. The library can then be further screened. High-performing variants can be isolated and further analyzed (e.g., by sequencing or chemical analysis) to identify them. Data from different variants can be input into a machine learning algorithm to further improve its predictive capabilities. The machine learning algorithm can also generate novel combinations of variants and / or application conditions that are expected to perform well. These variants and conditions can then be input into a strain optimization unit for further screening.
[0244] Figure 1 schematically illustrates the integrated system described herein. The integrated system 100 may be a platform comprising hardware units, consumption units, and biological or chemical units, each of which may be part of one or more integrated units described herein. The integrated system 100 may include integrated units such as a strain optimization unit 105, a molecular optimization unit 110, and a molecular discovery unit 115. Each integrated unit may be configured to take a library (e.g., of genetic or molecular variants) and a set of conditions (e.g., fermentation conditions, application conditions) as input and output a prediction (e.g., an optimized strain, an optimized molecule, or a novel molecule). The processes performed within each integrated unit may be recursive, as described herein.
[0245] The strain optimization unit 105 can be configured to take a strain optimization library containing multiple genetic variants as input, which can be generated using, for example, random mutagenesis, MAGE, CREATE, transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), CRISPR-guided DNA polymerase, error-prone polymerase chain reaction (epPCR), or a library from a database, such as a whole-genome sequencing (WGS) library. The strain optimization unit 105 can also take a series of fermentation conditions under which cells containing the genetic variants are grown as input. As output, the strain optimization unit 105 can generate an optimized strain (or information thereof) that has threshold performance for a specific application (e.g., enzyme activity) under a series of conditions.
[0246] The molecular optimization unit 110 can be configured to accept a molecular optimization library as input, which may contain multiple genetic variants encoding (or having sequence homology to) one or more target proteins. Genetic variants can be generated using techniques such as saturated mutagenesis, cyclic substitution, MAGE, CREATE, transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), CRISPR-guided DNA polymerase, and error-prone polymerase chain reaction (epPCR). In some examples (not shown), the molecular optimization library may contain multiple molecular variants screened for functional properties during the molecular optimization process. The molecular optimization unit 110 can also accept a set of applicable conditions as input, including but not limited to temperature, culture medium composition, oxygenation, salinity, and pH. As output, the molecular optimization unit 110 can generate optimized molecules (or information about them, e.g., DNA or protein sequences) that have threshold performance for specific applications (e.g., catalytic activity, molecular interactions, etc.) under a set of conditions.
[0247] The molecular discovery unit 115 can be configured to accept a molecular discovery library as input, which may contain multiple molecular variants. In some examples, the molecular variants may include multiple genetic variants that have known and unknown sequences. The molecular variants may arise from libraries collected from samples (e.g., biological samples, soil samples, etc.), or the libraries may be generated from, for example, metagenomic libraries, DNA coding libraries, function generator libraries, etc. Genetic variants may encode one or more proteins and / or have desirable functional properties. The molecular discovery unit 115 can also accept a set of applicable conditions as input, including but not limited to temperature, culture medium composition, oxygenation, salinity, pH, etc. As output, the molecular discovery unit 115 can generate novel molecules (or information thereof) that have threshold performance for specific applications (e.g., catalytic activity, molecular interactions, etc.) under a set of conditions.
[0248] Figure 2 schematically illustrates an integrated system for selecting a desirable biological variant. The integrated system 200 can be configured to accept multiple inputs 210, including a library input 211 (e.g., a strain library or DNA library), an assay input 213 (e.g., experimental data on the functional properties of members of the library), and a performance requirements input 215, which includes, for example, production conditions and performance traits (e.g., production or expression levels, substrate conversion, production under a range of environmental conditions, stability, etc.). The integrated system 200 can output one or more intermediate outputs 217 (e.g., between one or more cycles of a recursive cycle), which may include data from an integrated unit (e.g., an experimental module), such as sequence data, enriched strains, or enriched proteins. The intermediate outputs 217 may also include predictions about how a range of parameters can function on a larger scale, outputs of data including the functional performance of the inputs under specific conditions, and predictions of possible scale-up of performance. The intermediate outputs 217 can be reintegrated into the integrated system 200 as inputs. The final output 219 of the integrated system may include predicted and optimized strains, optimized molecules, or novel molecules expected to exhibit improved functional performance.
[0249] Computer system This disclosure provides a computer system programmed to carry out the methods of this disclosure. Figure 3 shows a computer system 301 that is programmed or otherwise configured to input one or more data streams, store one or more datasets in a data repository, set threshold parameters for data, and / or to instruct one or more integration units of the integration units to perform or iterate diversity screening. The computer system 301 can coordinate various aspects of the strain or molecule optimization or discovery of this disclosure, such as controlling one or more components of a strain optimization unit, a molecule optimization unit, or a molecule discovery unit, inputting data, and outputting one or more predictions regarding structure-function relationships of molecules. The computer system 301 may be a computer system located on a user's electronic device or remotely located to an electronic device. The electronic device may be a mobile electronic device.
[0250] The computer system 301 includes a central processing unit (CPU, "processor" and "computer processor" as used herein) 305, which may be a single-core or multi-core processor, or multiple processors for parallel processing. The computer system 301 also includes memory or memory locations 310 (e.g., random-access memory, read-only memory, flash memory), an electronic storage unit 315 (e.g., a hard disk), a communication interface 320 for communicating with one or more other systems (e.g., a network adapter), and peripheral devices 325 such as a cache, other memory, data storage, and / or an electronic display adapter. The memory 310, storage unit 315, interface 320, and peripheral devices 325 communicate with the CPU 305 via a communication bus (solid line), such as a motherboard. The storage unit 315 may be a data storage unit (or data repository) for storing data. The computer system 301 can be operably coupled to a computer network ("network") 330 with the help of the communication interface 320. Network 330 can be the Internet, the Internet and / or an extranet, or an intranet and / or extranet communicating with the Internet. Network 330 may, in some cases, be a telecommunications and / or data network. Network 330 may include one or more computer servers that can enable distributed computing, such as cloud computing. Network 330 may, in some cases, implement a peer-to-peer network with the help of computer system 301, thereby enabling devices coupled to computer system 301 to act as clients or servers.
[0251] The CPU 305 can execute a sequence of machine-readable instructions, which can be embodied in a program or software. Instructions can be stored in a memory location, such as memory 310. Instructions can target the CPU 305, which can then be programmed or configured to perform the methods of this disclosure. Examples of operations performed by the CPU 305 may include fetching, decoding, executing, and writing back.
[0252] The CPU 305 can be part of a circuit, such as an integrated circuit. One or more other components of system 301 can be included in the circuit. In some cases, the circuit is an application-specific integrated circuit (ASIC).
[0253] The storage unit 315 can store files such as drivers, libraries, and saved programs. The storage unit 315 can also store user data, such as user preferences and user programs. The computer system 301 may, in some cases, include one or more additional data storage units located outside the computer system 301, such as on a remote server that communicates with the computer system 301 via an intranet or the internet.
[0254] Computer system 301 can communicate with one or more remote computer systems via network 330. For example, computer system 301 can communicate with a user's remote computer system. Examples of remote computer systems include personal computers (e.g., portable PCs), slate or tablet PCs (e.g., Apple® iPad®, Samsung® Galaxy Tab), telephones, smartphones (e.g., Apple® iPhone®, Android® compatible devices, Blackberry®), or personal digital assistants. Users can access computer system 301 via network 330.
[0255] The methods described herein can be implemented by machine-executable code (e.g., a computer processor) stored in an electronic storage location of a computer system 301, such as memory 310 or an electronic storage unit 315. The machine-executable code or machine-readable code can be provided in the form of software. In use, the code can be executed by the processor 305. In some cases, the code can be retrieved from the storage unit 315 and stored in memory 310 for easy access by the processor 305. In some situations, the electronic storage unit 315 can be omitted, and machine-executable instructions are stored in memory 310.
[0256] The code can be pre-compiled and configured for use in a machine with a processor adapted to run the code, or it can be compiled at runtime. The code can be supplied in a programming language that can be chosen to allow the code to run either pre-compiled or compiled.
[0257] Embodiments of systems and methods provided herein, such as computer system 301, can be embodied in programming. Various embodiments of the art can typically be considered “products” or “manufactured goods” in the form of machine (or processor) executable code and / or associated data contained on or embodied thereon on some kind of machine-readable medium. Machine executable code can be stored in memory (e.g., read-only memory, random-access memory, flash memory) or electronic storage units such as hard disks. The “storage” type medium can include any or all of tangible memory such as computers and processors, or associated modules such as various semiconductor memories, tape drives, disk drives, etc., which can provide non-temporary storage for software programming at any time. All or part of the software may be communicated over the Internet or various other telecommunication networks. Such communication can enable, for example, the loading of software from one computer or processor to another computer or processor, for example, from a management server or host computer to an application server computer platform. Therefore, other types of media on which software elements can be carried include optical waves, electrical waves, and electromagnetic waves, such as those used across physical interfaces between local devices, via wired and optical terrestrial networks, and via various air links. The physical elements that carry such waves, such as wired or wireless links and optical links, can also be considered media on which software can be carried. As used herein, unless limited to non-temporary and tangible “storage” media, terms such as computer or machine “readable media” refer to any medium involved in providing instructions to a processor for execution.
[0258] Therefore, machine-readable media such as computer executable code can take many forms, including but not limited to tangible storage media, carrier media, or physical transmission media. Non-volatile storage media include optical or magnetic disks, such as any storage device, such as a computer, that can be used to implement a database, for example, as shown in the drawing. Volatile storage media include dynamic memory, such as the main memory of a computer platform. Tangible transmission media include copper wires, such as coaxial cables, and optical fibers, including wires that have buses within a computer system. Carrier media can take the form of electrical or electromagnetic signals, or sound or light waves, such as those generated during radio frequency (RF) and infrared (IR) data communications. Therefore, common forms of computer-readable media include, for example, floppy disks, flexible disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, DVDs or DVD-ROMs, any other optical media, punch cards, paper tapes, any other physical storage media having a pattern of holes, RAM, ROMs, PROMs and EPROMs, FLASH-EPROMs, any other memory chips or cartridges, carriers that carry data or instructions, cables or links that carry such carriers, or any other media from which a computer can read programming code and / or data. Many of these forms of computer-readable media can be involved in transporting one or more sequences of one or more instructions to a processor for execution.
[0259] The computer system 301 includes, or can communicate with, an electronic display 335, which includes, for example, a user interface (UI) 340 for providing input threshold parameters for selecting a desired biological variant. Examples of UIs include, but are not limited to, graphical user interfaces (GUIs) and web-based user interfaces.
[0260] The methods and systems of this disclosure can be implemented by one or more algorithms. The algorithms can be implemented by software during execution by the central processing unit 305. The algorithms may be, for example, machine learning algorithms that iteratively input a data stream relating to biological diversity or conditional diversity and output predictions relating to the structure-function relationships of molecules, optimized strains, optimized molecules, or discovered molecules. [Examples]
[0261] Example 1: Stock Optimization Matching microscale screening conditions precisely to those experienced at larger scales can be challenging. Furthermore, the physical limitations of mixing in large-scale containers can result in strains in highly heterogeneous environments, where cells are exposed to gradients in conditions such as oxygen levels, pH, carbon source concentration, and temperature. In conventional screening programs, mutant screening can often be a time and cost bottleneck, allowing for the selection of specific screening conditions that are closest to large-scale production conditions. Matching these conditions can be difficult, and the heterogeneity of conditions at large scales may make it less likely that strains selected to improve from a particular small-scale screening assay will actually function well at larger scales.
[0262] Miniaturizing the assay can reduce screening time and cost to a degree that does not significantly hinder screening large populations under various conditions. In this example, a library is screened independently under multiple conditions, and next-generation sequencing (NGS) is used to identify mutations enriched under these different conditions. The mutations are categorized into metabolic pathways they affect. This phenotypic, sequence, and metabolic pathway network data is then fed into a machine learning algorithm to enable prediction of which mutations, pathways, or combinations are most important for performance under different environmental conditions, and which mutations provide robust improvements across multiple conditions and are therefore generally useful.
[0263] Stock construction Starting with the Thermo-Fisher protease-deficient commercial strain pichia pastoris SMD1168, the human type I collagen gene COL1A1 was PCR-amplified and cloned into the pPIC9K plasmid, followed by transformation of the pichia pastoris SMD1168 strain with the resulting plasmid. The pPIC9K plasmid contains a secretion tag that enables the generation of a multicopy plasmid under His4 selectivity and allows for extracellular production of the protein.
[0264] Building a stock optimization library The library is created by whole-genome mutagenesis of modified strains. Logarithmic phase cells are then exposed to N-methyl-N'-nitro-N-nitroguanidine (NTG) and quenched with sodium thiosulfate at appropriate concentrations and times to obtain a viability of 0.5–1%.
[0265] Microscale assay for collagen production To detect collagen production from piquiya cells, a PDMS nanowell array is constructed (Torres AJ, Hill AS, Love JC. Nanowell-based immunoassays for measuring single-cell secretion: characterization of transport and surface binding. Anal. Chem. 2014;86(23):1152-11569. doi:10.1021 / ac4030297). Cells from the library are suspended in growth medium containing 0.5% methanol to induce protein expression. Cells are distributed into PDMS nanowells so that approximately one cell is present in 10 wells to ensure single-cell measurement. The wells are covered with glass slides coated with anti-COL1A1 antibody clone 5D8-G9 (Millipore-Sigma). Collagen secreted by the cells is captured by the antibody on the glass surface. The glass slides are then removed, stained with FITC-conjugated goat anti-mouse antibody, and imaged with a fluorescence microscope. FITC intensity is proportional to the amount of collagen secreted by the cells. Nanowells containing highly expressive clones are identified, extracted from the nanowell array using a robotic micromanipulator, and then grown under non-inducible conditions for further study.
[0266] Stock improvement Multiple screening experiments will be set up in parallel to cover a range of conditions that cells are likely to experience in production. For each condition, the most productive cells will be selected, regrow, and the improvement in production on a 1 ml scale in a 96-well plate will be confirmed. After performance confirmation, DNA will then be extracted and next-generation sequencing (NGS) will be performed on either pooled DNA from each condition or individual clones, using Illumina's MiSeq, for example. Mutations associated with production under different conditions will be identified.
[0267] The strains found in the library are unlikely to be perfectly optimized. To eliminate unhelpful mutations and discover beneficial combinations, protoplast fusion and "genome shuffling" are used to enable recombination between genomes of high-productivity strains. The recombinant library is screened under the same initial conditions used to select the highest-producing strains.
[0268] The library creation, screening, and analysis cycle is repeated recursively until a strain that meets the performance requirements is identified.
[0269] Example 2: Molecular discovery Discovery of novel lipase enzymes from a metagenomic library. Environmental DNA (eDNA) is collected and purified from soil or other sources. The eDNA is cut and size-selected to the range of 2–5 kb to include typical gene and promoter sizes. The fragments are cloned into plasmids and then transformed into highly transformable E. coli.
[0270] Lipase assay E. coli cells are combined with the commercially available fluorescent substrate EnzCheck Lipase Substrate (ThermoFischer) and cell lysis reagent B-PER (ThermoFischer). The mixture is emulsified by shaking with a tissue lysis homogenizer (TissueLyser II, Qiagen) using HFE-7500 (3M), a mixture of surfactant and oil containing 2% "008" fluorosurfactant (RAN Biotechnologies). The emulsion is transferred to a shallow flow cell with a punctureable, light-permeable membrane. A 10 × 50 × 0.25 mm flow cell can store approximately 125 μL of emulsion, corresponding to an emulsion compartment of approximately 108 picoliters. When enzymes are released from the lysed cells, droplets containing cells expressing active lipase become fluorescent. The flow cell is imaged using a fluorescence microscope, and the fluorescent area is collected by puncturing the coating membrane and aspirating with a micromanipulator needle.
[0271] Plasmids enriched with the gene encoding the active lipase enzyme are isolated and retransformed into E. coli. The transformation, assay, and harvesting cycle is repeated until the population is substantially enriched with the active lipase gene.
[0272] Next, individual clones are isolated from the selection and grown separately. Lipase activity is confirmed, and the inserts are sequenced to identify the open reading frames encoding the lipase. The sequences are checked for homology with known lipase gene families to identify new classes. There are two pathways depending on the number of unique clones remaining in the library after enrichment. If the number is between 1 and approximately 100, the enzymes are evaluated individually for activity. In contrast, if the number of unique clones exceeds 100, further characterization is performed at the population level.
[0273] Characteristic evaluation To characterize sequences at the population level, the output from the above activity enrichment, containing hundreds to thousands of unique active enzymes, is used. This same pool is subjected to high-throughput emulsion-based enrichment as described above multiple times in parallel under several different conditions. In this multi-condition characterization, performance is evaluated across several key parameters, including substrate state (solubility and surface adsorption), stability and catalytic performance at low and high temperatures, resistance to protease degradation, and sensitivity to divalent cation concentration. Enrichment is performed under low and high conditions for each of these parameters, and the population is enriched for the best performance under each condition. The differential enrichment of sequences across these parameters is then measured using long-read NGS (LoopSeq®, LOOP Genomics). These differences in enrichment correlate with the performance of these variants in the assay.
[0274] Array data, structure inference, and performance data are used to construct a machine learning training set that can be used to predict which regions, motifs, and structural classes function well under different conditions. This model is used as an input to drive molecular optimization.
[0275] Example 3: Molecular Optimization - Optimization of Lipase Enzymes for Activity at Low Temperatures The output of the molecular discovery and characterization unit contains information regarding sequence identity and performance. Data is obtained regarding which enzyme families, regions within the enzyme, and specific amino acid residues are associated with higher low-temperature performance. An existing sequence database is searched for natural variants predicted to have good performance based on an algorithm. This information is used to construct a lipase gene library targeting these regions. The library is transformed into a suitable host and screened for those having the highest performance at a target temperature of 15°C. If the library diversity is less than 1,000, screening is performed in a conventional microtiter format, and if the diversity is large, screening is performed with microfluidic plugs or droplets.
[0276] In either case, protein variants are expressed in bacteria, and then lysates of these bacteria are exposed to a fluorescently generating substrate. The population is enriched for variants having the highest activity at low temperature. The enriched population is then sequenced and compared to the starting population. Both the positively enriched sequences and the negatively enriched sequences result in a model that enables prediction of the best-performing enzymes in our system. This data is used to design a partial sequence library in which important residues affecting performance are saturated. The highest-performing sequences are also recombined by gene shuffling or synthesis of promising combinations. The cycle of enrichment and library improvement is repeated until the performance goal is reached.
[0277] Example 4: Strain Optimization by Random Mutagenesis The methods, compositions, and systems of this specification can enable strain optimization in a high-throughput format with high predictability. In one example, a strain expressing a molecule of interest can be subjected to non-target mutagenesis, such as random mutagenesis. Random mutagenesis can involve subjecting a strain expressing a molecule of interest to a mutagen and / or random genetic changes (e.g., transposon mutagenesis, MAGE, genomic shuffling, random recombination, non-homologous end joining, etc.), thereby generating a library of mutagenized cells. Such a library can contain, for example, more than 10,000 variants or more than 1,000,000 variants. A library of mutagenized cells (considered one generation) can be expanded by growth in a liquid culture medium, and the cells can be processed to ensure that the population consists of single cells.
[0278] Figure 4 schematically shows an exemplary workflow for performing strain optimization. In process 401, as described above, a library of variants (or mutagenized cells, also referred to herein as genetic variants) can be generated. In process 4, the library of variants can be distributed to individual reaction vessels (e.g., microreactors). In some examples, individual cells or single cells are distributed to individual vessels. The distribution may be assisted by a computer-implemented method having a feedback loop that ensures that all individual vessels contain a predetermined volume (within a range), which can be useful for ensuring consistent growth among the individual cells within the individual vessels. In some examples, image recognition can be used to ensure that a desired number of cells are loaded into the individual vessels.
[0279] Individual containers containing individual cells can be subjected to conditions that mimic fermentation conditions (e.g., temperature, carbon source, pH, culture medium components, aeration, etc.). Individual containers may contain the same set of fermentation conditions, or different fermentation conditions, or a subset of individual containers may contain the same set of fermentation conditions, or another subset of individual containers may contain different sets of fermentation conditions. Cells in individual containers can be incubated within the individual container for a timeframe that allows for cell proliferation and / or production of the desired molecule. Growth within the microbioreactor can be monitored by imaging and computer image recognition.
[0280] In process 405, individual containers can be screened, and based on the screening, one or more cells (of a series of distributed mutagenic cells) can be enriched (e.g., improved cells or cells expressing a desired phenotype can be selected). In one example, individual containers or their contents can be placed in a device that allows for the addition of reagents and control of reaction time. Detection of the target molecule can be performed, and based on the detection, one or more cells can be selected. Detection and / or selection can be performed in an automated manner. For example, the system described herein may include a computer-aided method that can determine which individual containers contain cells exhibiting the target phenotype by examining molecular generation and molecular activity via the addition of a detection reagent. The environment of the target molecule (e.g., pH, ionic strength, ionic type, substrate, inhibitor, enzyme, competitor) can be altered using the addition of a detection reagent. Such environmental changes may be useful in generating conditions that reflect the amount of molecule produced and / or the function of the molecule produced. Individual containers can be evaluated for the target molecule via a spectroscopic signal at a predetermined controllable time after the addition of the reagent. Instrument design and computer control feedback for instruments can control decision-making regarding timing, reagent addition, molecular generation evaluation, and which microbioreactors meet selection criteria. In some cases, PID algorithms can be used to prevent erroneous selection decisions caused by instrument drift during the strain optimization process (e.g., due to slight deviations in mechanical instruments, liquid viscosity, temperature, etc.).
[0281] If necessary, after selection, for example when determining the instrument for holding specific library members (e.g., mutagenic cells), the selected cells can be removed from their container by the system and held in an environment that maximizes the viability of the organisms (e.g., sufficient nutrient conditions, salt conditions, temperature, aeration, etc.). The cells can be expanded through cell proliferation, and if necessary, screening and enrichment can be carried out in other processes (e.g., processes 403 and 405, 407 in some examples). If necessary, more stringent conditions can be applied to each screening and enrichment performed (e.g., lower oxygen, lower nutrients, pH changes, etc.). Alternatively or additionally, more stringent performance criteria (e.g., increased expression, catalytic activity, etc.) can be applied during enrichment. The enriched cells can then be characterized in process 407. The enriched cells can be analyzed, for example, through sequencing or phenotypic screening. For example, the enriched cells can be analyzed to confirm the desired phenotype under further test conditions. Cells retaining the desired phenotype are then introduced into another iteration ("generation") of genetic alteration and screening (e.g., processes 401, 403, and 405), and the process can be repeated in recursive cycles until performance criteria are met.
[0282] Figure 5 shows an exemplary workflow of the strain optimization process described herein. The parent strain or cell line 501 is subjected to random mutagenesis (9 exposures to ultraviolet light) to provide a first library of mutants, e.g., mutagenic cells 503. The first library of mutagenic cells 503 is subjected to enrichment in a first process, which may include distribution, optional proliferation, and sorting. Mutants with functional performance exceeding a threshold can be collected and then subjected to enrichment in another process to produce a second library of mutagenic cells 505. The enrichment process can be carried out for any number of useful processes to obtain subsequent libraries of mutagenic cells (e.g., 507 and 509). The final library of mutagenic cells can then be collected and characterized (e.g., via sequencing) to perform additional mutagenesis and initiate a new generation, as described herein, for example. The new generation of parent strain or cell line 511 can be further subjected to mutagenesis (exposure to ethyl methyl sulfonate (EMS)) to perform one or more enrichments, and the process can be repeated until an optimal or desired performance criterion is met.
[0283] Figures 6A–6C show exemplary data obtained by performing strain optimization over three generations (e.g., three cycles of mutagenesis, distribution, enrichment, and characterization). Figure 6A shows histograms of fluorescence intensity (x-axis) that may represent the amount or quantity of the target molecule (e.g., enzyme) for individual cells before strain optimization 601, which may not have been mutageneised, and after the first enrichment process 603 and the second enrichment process 605 within the first generation. The highest-performing variant from the final enrichment process of the first generation (e.g., the variant producing the maximum quantity of the target molecule) is then further mutagenerated, thereby initiating the generation of the second generation of mutants. Figure 6B shows histograms of fluorescence intensity (x-axis) for individual cells before strain optimization 607 in the second generation, after the first process 609, after the second process 611, and after the third process 613. Figure 6C shows histograms of the fluorescence intensity (x-axis) of individual cells before strain optimization 615 in the third generation, after the first process 617, after the second process 619, and after the third process 613. As shown in Figures 6A to 6C, the additional enrichment process results in an increase in intensity (right shift), indicating improved performance.
[0284] Figure 7 shows an exemplary plot of the concentrations of products generated from mutants obtained over two generations of mutagenesis and multiple selection and enrichment cycles, comparing cultivation in a reaction vessel (e.g., used for high-throughput screening and selection) with cultivation in a bulk-feed batch reactor (y-axis) (x-axis). Each point represents the amount (concentration) of product generated from a single mutant (the most expressive mutant) from each generation when grown under both microreactor and batch growth conditions. As shown in Figure 7, the generation of the target molecule (e.g., protein) in the microreactor is nearly linearly proportional to the generation of the target molecule (e.g., protein) at the batch scale, thus demonstrating that the strain optimization process can yield results representative of batch growth conditions at the microscale. Overall, these data demonstrate how strain optimization can be an effective tool for improving the functional performance of strains in a high-throughput manner.
[0285] Example 5: Strain optimization using a natural diversity library The methods, compositions, and systems described herein can enable strain optimization in a high-throughput format with high predictability. For example, a strain expressing a molecule of interest can be subjected to genetic alteration based on a large set of natural and / or manipulated diversity. The diversity set may include promoters, 5' untranslated regions, ribozymes, RNA-stable sequences, secretory peptides, signal peptides, fusion proteins, DNA-binding domains, protein-protein-binding regions, codon-optimizing genes, codon-randomization genes, 3' untranslated regions, terminators, non-homologous end joins, etc., which can be introduced into multiple cells to generate a library of mutagenic cells. The library of mutagenic cells may include, for example, more than 1,000 variants, more than 10,000 variants, more than 100,000 variants, or more than 1,000,000 variants. The library of mutagenic cells (considered to be one generation) can be expanded by growth in a liquid culture medium, and the cells can be processed to ensure that the population consists of single cells.
[0286] To initiate the process within a generation, individual cells from a library of mutagenic cells can be distributed into individual containers, subjected to fermentation conditions, monitored as needed, and screened as described above (see Example 4) and as shown in Figures 4-5.
[0287] Figure 8 shows exemplary data obtained by performing strain optimization over a single generation (e.g., one cycle of mutagenesis). In Figure 8, each panel shows a histogram plot of intensity representing the quantity of the target molecule produced. Non-mutagenic cells 803 (e.g., cells before mutation introduction) maintain relatively similar peak profiles as expected, but the mean of the entire population of mutagenic cells 801 obtained from each process (e.g., mutagenic cells that are distributed, cultured, and screened) indicates that the production of the target molecule is initially lower, possibly due to the adverse effects of mutagenesis. Over the subsequent screening process (e.g., distribution, enrichment, and any characterization), the mean of the entire population of mutagenic cells 801 shows a peak shift to the right, which indicates an increase in intensity, and therefore the quantity of the target molecule produced, as a result of the strain optimization and screening processes.
[0288] Figure 9 shows the results of a plate-based assay for measuring the target molecule across each screening process within a single generation. Each bar represents a measured value of the multiplicative change in intensity (representing the quantity of the target molecule produced) across the strain optimization process, normalized to non-mutagenic cell generation 903. Mutagenic cells 901 obtained from each process (each set of bars) (e.g., mutagenic cells that are distributed, cultured, and screened) show increased intensity per step, and therefore a greater quantity of the target molecule produced as a result of the strain optimization process.
[0289] Example 6: Molecular discovery by targeted or untargeted mutagenesis The methods, compositions, and systems described herein can enable molecular discovery, for example, the discovery of genetic variants that result in phenotypic improvements. For example, cells can be subjected to mutagenesis (e.g., targeted or untargeted mutagenesis) to produce a library of mutagenic cells. Such a library may contain, for example, more than 10,000 variants or more than 1,000,000 variants. The library of mutagenic cells can be expanded by growth in a liquid culture medium, and the cells can be processed to ensure that the population consists of single cells.
[0290] Figure 10 schematically illustrates an exemplary workflow for performing molecular discovery. The process may be broadly similar to the process shown in Figure 4. In process 1001, a library of mutants (or mutagenic cells, also referred to herein as genetic variants) can be generated, for example, using targeted or untargeted mutagenesis, as described above. In process 1003, the library of mutants can be distributed into individual containers (e.g., microreactors). In some examples, individual cells or single cells are distributed into individual containers. Distribution may be assisted by a computer implementation method having a feedback loop that ensures all individual containers contain a predetermined volume (within a range), which may be useful in ensuring consistent growth among individual cells within individual containers. In some examples, image recognition can be used to ensure that a desired number of cells are filled into individual containers.
[0291] Individual containers containing individual cells can be subjected to conditions that mimic fermentation conditions (e.g., temperature, carbon source, pH, culture medium components, aeration, etc.). Individual containers may contain the same set of fermentation conditions, or different fermentation conditions, or a subset of individual containers may contain the same set of fermentation conditions, or another subset of individual containers may contain different sets of fermentation conditions. Cells in individual containers can be incubated within the individual container for a timeframe that allows for cell proliferation and / or production of the desired molecule. Growth within the microbioreactor can be monitored by imaging and computer image recognition.
[0292] In process 1005, individual containers can be screened, and based on the screening, one or more cells (of a series of distributed mutagenic cells) can be enriched (e.g., improved cells or cells expressing a desired phenotype can be selected). In one example, individual containers or their contents can be placed in a device that allows for the addition of reagents and control of reaction time. Detection of the target molecule can be performed, and based on the detection, one or more cells can be selected. Detection and / or selection can be performed in an automated manner. For example, the system described herein may include a computer-aided method that can determine which individual containers contain cells exhibiting the target phenotype by examining molecular generation and molecular activity via the addition of a detection reagent. The environment of the target molecule (e.g., pH, ionic strength, ionic type, substrate, inhibitor, enzyme, competitor) can be altered using the addition of a detection reagent. Such environmental changes may be useful in generating conditions that reflect the amount of molecule produced and / or the function of the molecule produced. Individual containers can be evaluated for the target molecule via a spectroscopic signal at a predetermined controllable time after the addition of the reagent. Instrument design and computer control feedback for instruments can control decision-making regarding timing, reagent addition, molecular generation evaluation, and which microbioreactors meet selection criteria. In some cases, PID algorithms can be used to prevent erroneous selection decisions caused by instrument drift during the strain optimization process (e.g., due to slight deviations in mechanical instruments, liquid viscosity, temperature, etc.).
[0293] If necessary, after selection, for example when determining the instrument for holding specific library members (e.g., mutagenic cells), the selected cells can be removed from their container by the system and held in an environment that maximizes the viability of the organisms (e.g., sufficient nutrient conditions, salt conditions, temperature, aeration, etc.). The cells can be expanded through cell proliferation, and if necessary, screening and enrichment can be carried out in other processes (e.g., processes 1003 and 1005). If necessary, more stringent conditions can be applied to each screening and enrichment performed (e.g., lower oxygen, lower nutrients, pH changes, etc.). Alternatively or additionally, more stringent performance criteria (e.g., increased expression, catalytic activity, etc.) can be applied during enrichment. The enriched cells can then be characterized in process 1007. The enriched cells can be analyzed, for example, through sequencing or phenotypic screening. For example, the enriched cells can be analyzed to confirm the desired phenotype under further test conditions.
[0294] In some cases, the characterization of process 1007 involves sequencing of genes and / or proteins. For example, enriched cells can be subjected to DNA or RNA sequencing to determine or identify the presence of one or more mutations, or to characterize the mutations. In some cases, the amino acid sequence of the resulting transcription or translation peptide of the DNA or RNA sequence can be determined based on the DNA or RNA sequence. The amino acid sequence can be compared to the amino acid sequence of the parent (e.g., non-mutagenic) cell to determine specific changes in the amino acid sequence of the protein of interest (e.g., which residues of the peptide are different). Further validation studies can be carried out, for example, by transforming or transducing parent (e.g., non-mutagenic) cells or daughter cells (e.g., mutagenic cells derived from parent cells) with the gene sequence of the enriched variant. For example, process 1009, in which variants are enriched, their nucleic acid or protein sequences are identified, and then those variants are transferred into cells (e.g., via transformation or transduction) so that the cells contain the identified nucleic acid or protein sequence of the enriched variant, can result in cells with improved phenotype or functional performance.
[0295] Figure 11 shows exemplary data from performing molecular discovery using the workflow described above and shown in Figure 10. Molecular discovery is performed for a single generation (e.g., one cycle of mutagenesis) over two distribution, enrichment, and optional characterization. In Figure 11, each panel shows an intensity histogram representing the quantity of the target molecule (e.g., secreted protein) produced. In one example, the target molecule produced may be a reporter protein (e.g., green fluorescent protein) whose expression can be driven by a SigA promoter. Higher SigA activity results in the expression of more reporter protein and a higher signal intensity. Non-mutagenic cells 1103 (e.g., cells before mutation introduction) maintain relatively similar peak profiles, but mutagenic cells 1101 obtained from each process (e.g., mutagenic cells that are distributed, cultured, and screened) show a shift to the right, indicating an increase in intensity and therefore the quantity of the target molecule produced as a result of the molecular discovery process. The panel below shows mutagenic cells 1101 after a second gating and sorting, demonstrating that an increase in the desired phenotype (e.g., higher intensity, indicating a greater quantity of the desired molecule produced) is observed after sorting and enrichment.
[0296] Following or simultaneously with screening, a population of mutagenic cells can be selected. For example, the system described herein may include a computer-readable medium for setting threshold functional performance parameters (e.g., minimum intensity, minimum concentration, minimum quantity of the molecule of interest) from which cells are selected. Cells exceeding the threshold may be selected or concentrated in a first bin for further analysis, while cells below the threshold may be selected in a second bin. Cells in the first bin can be further analyzed for mutations, for example, to determine nucleic acid or protein sequences, which may help determine which mutations result in improved functional performance.
[0297] Figure 12 shows a table of exemplary mutations obtained from molecular discovery processes, e.g., the process outlined in Figure 10. Each mutation is in the form XNNNY, where X represents the original amino acid residue (e.g., the residue in a non-mutant cell), NNN represents the amino acid position of the mutation, and Y represents the original amino acid residue to which it was mutated. For example, "T31R" refers to a mutation from threonine to arginine at the 31st amino acid residue. Each mutation in Figure 12 is for the SigA transcription factor, which can modify the expression of the target protein or reporter protein (e.g., GFP). The mutants shown in Figure 12 result in higher expression of the target protein or reporter protein. Mutations have been demonstrated to result in improved functional performance (e.g., increased secretion or activity of the SigA transcription factor, and therefore increased expression and intensity of the reporter protein), which is a result of enrichment during the molecular discovery process.
[0298] Figure 13 shows exemplary data of introducing the induced mutations of Figure 12 into cells. In such examples, the cells may be non-mutagenic cells, parental cells, or mutagenic cells. The cells can be transformed with a gene variant that expresses the protein variant identified in Figure 12 (e.g., a SigA variant that can promote the expression of the protein of interest or a reporter protein such as GFP). The expression of the protein variant (e.g., a SigA variant) can be controlled by an inducible promoter. Panel A of Figure 13 shows a control group in which the parent (e.g., a non-mutagenic cell) can be transformed with a plasmid vector control having an inducible promoter. The induction of expression (e.g., by turning on the inducible promoter, shown in this example as the addition of isopropyl β-d-1-thiogalactopyranoside, IPTG) does not result in a significant difference in the fluorescence signal of the reporter protein, indicating little to no change in the expression or activity of the transcription factor (e.g., SigA), as expected. In Panel B, cells expressing a reporter protein (e.g., GFP, whose expression is driven by SigA) that naturally produces a transcription factor (e.g., SigA) are transformed with a plasmid containing an inducible gene encoding the transcription factor (e.g., SigA). Forced expression of the wild-type transcription factor (e.g., SigA), already naturally produced by the cells, results in increased reporter protein signaling and indicates some increase in the quantity or activity of the transcription factor (e.g., SigA). In Panels C–F, the expression of the reporter protein (e.g., GFP) is increased for each of the introduced induced SigA mutants. In each of Panels C–F, cells are transformed with plasmids containing one or more of the enriched SigA mutants listed in Figure 12. Each of the mutant groups (Panels C–F) increases reporter protein expression by 2.5-fold, 3.7-fold, 2.4-fold, and 3.7-fold, respectively, upon induction of the inducible promoter. Overall, these results indicate that the molecular discovery systems and methods described herein may be useful for identifying genetic variants that result in improved functional performance, such as increased protein expression or activity.
[0299] Example 7: Improvement of Molecules Using Machine Learning The methods, compositions, and systems of this specification can enable the improvement or optimization of molecules, such as the generation of improved molecular variants. In one example, the methods of this specification can be used to produce improved polymer variants, such as peptides or proteins.
[0300] FIG. 14 schematically shows an exemplary workflow for performing molecule improvement to produce improved polymer variants (e.g., proteins). In process 1401, a library of polymer variants is generated. In one example, the library of polymer variants can include a library of variants of a gene encoding an enzyme, and such a library of variants can be generated by mutagenesis. For example, for a given gene, mutagenesis can be performed to obtain a site-saturation library (SSL) in which all amino acid positions in the protein have been changed to all 20 amino acids. Further mutagenesis can be performed, for example, via gene shuffling and error-prone PCR to obtain a library of variants. The library of variants can then be transformed into a plurality of cells (e.g., a suitable bacterial expression host) such that each cell contains a single variant (e.g., a sequence encoding a polymer, peptide, or protein variant).
[0301] In process 1403, the mutant library can be distributed into individualized containers. If necessary, the containers may contain conditions sufficient for cells to grow and produce the polymer of interest (e.g., protein or enzyme). After several hours of growth and production, the cells can then be screened and enriched in process 1405. During screening, enzyme substrates in reaction buffer can be added to each individualized container. The individual containers are incubated for a precise time, and then the function of the polymer variants can be determined to obtain functional data. For example, enzyme activity may be measured (e.g., by measuring the amount of fluorescence or luminescence product). Simultaneously with or following the measurement, the mutant library can be sorted into several functional gates (e.g., four gates) representing different performance (e.g., enzyme activity). In some examples, a subset of the mutant library may not be sorted and may be excluded as a control population or comparison (e.g., pre-enrichment) group.
[0302] In some cases, the total number of cells measured may be greater than the number of unique polymer variants. For example, after proliferation and culture of variants (e.g., mutagenic cells containing genetic variants), each variant may contain a clonal population of variants. After distributing single cells into individual reaction vessels, multiple copies of the same variant may exist within each individual reaction vessel. Therefore, the total number of cells assayed may be 50 to 100 times the number of unique variants expected in the library.
[0303] In process 1407, for example, the selected cells from one or more of the selected gates can be further analyzed. For example, each cell (or cell population) can be isolated from the container and sequenced using next-generation sequencing to obtain sequence data. Unselected cells can also be sequenced. Based on the sequence data, the abundance of each variant can be calculated both before enrichment (from unselected cells) and after enrichment (from selected cells). The enrichment data for each sequence, e.g., the enrichment factor, can be calculated by comparing the abundance of each variant in the selected cells with the abundance of each variant in the unselected cells. Since the enrichment factor is proportional to the functional performance of the variant (e.g., catalytic activity), the enrichment factor can be used as a functional label for each sequence in the library. This sequence / functionally labeled dataset can then be input and / or analyzed in process 1409 using machine learning (ML) techniques to make recommendations for future designs. The ML algorithm can generate thousands of designs (e.g., genetic variants or polymer variants) that are predicted to have high performance. These predicted variants can then be synthesized (e.g., using oligonucleotide arrays) and assembled into plasmids for transformation into an expression host. The molecular improvement or optimization process (e.g., processes 1401, 1403, 1405, 1407, and 1409) can then be repeated several times (e.g., 2–4 generations) to obtain highly active variants.
[0304] Figure 15 shows an example of data from the molecular optimization process. Each panel shows a histogram plot of intensity representing the performance (e.g., enzyme activity level) of the target molecule generated (from top to bottom) over the subsequent enrichment process. Non-mutagenic cells 1503 (e.g., cells before mutation introduction) maintain relatively similar peak profiles as expected, but mutagenic cells 1501 obtained from each process (e.g., mutagenic cells that are distributed, cultured, and screened) initially show lower performance (e.g., activity level) of the target molecule, likely due to the adverse effects of mutagenesis. Over the subsequent screening process (e.g., distribution, enrichment, and any characterization), mutagenic cells 1501 show a peak shift to the right, which indicates an increase in intensity and, therefore, the activity level of the target generated molecule, as a result of the molecular optimization process.
[0305] Figure 16 shows a scatter plot of variant activity across generations of molecular optimization. As can be seen from Figure 16, variant activity (x-axis) increases over several generations.
[0306] Example 8: Equipment Control This specification also describes instrument control algorithms that may be useful for compensating for drift, noise, and / or creep that occur within the system or during the methods described herein.
[0307] In some examples, the computer-readable medium may include one or more feedback algorithms for controlling the distribution frequency (e.g., controlling the frequency of droplets), which may include automatically adjusting, for example, flow rate, pressure, temperature, etc. In another example, the sorted gate, e.g., threshold parameters, may be adjusted to account for noise, creep, etc. For example, a threshold gate (e.g., a minimum intensity threshold or maximum percentage of the library) may be set during sorting or concentration. The feedback algorithm may automatically change or modulate the gating parameters to maintain the collection ratio.
[0308] Figure 17 shows exemplary data for a sorting gate feedback system. The upper plot shows the signal intensity of the measured variants as a product of time. The upper line 1701 and the lower line 1703 represent the gates. As the signal diverges higher over time (drifts over time), the feedback algorithm adjusts the two gates 1701 and 1703 to maintain a constant percentage of enriched variants between the gates. The lower plot shows the percentage of variants within the collection range as a function of time, which remains relatively constant.
[0309] Embodiment In some cases, this disclosure provides methods according to the following embodiments. Embodiment 1. A method for producing an improved polymer variant, (a) obtaining a dataset, i) Sequence data including polymer variant sequence data, ii) Obtaining enrichment data based on the abundance of polymer variant sequences in samples selected based on polymer variant function, (b) A method comprising applying machine learning algorithm 1 or more to the dataset to design improved polymer variants. Embodiment 2. The method according to Embodiment 1, wherein the polymer variant sequence includes i) the sequence in an unselected sample and ii) the sequence in the sample selected based on the polymer variant function, and the sample selected based on the polymer variant function and the unselected sample originate from a common sample. Embodiment 3. The method according to Embodiment 2, wherein the concentration data is further based on the amount of the polymer variant sequence present in the unselected sample. Embodiment 4. The method according to Embodiment 3, wherein the amount of the polymer variant sequence in the sample selected based on the polymer variant function includes sequence counting. Embodiment 5. The method according to Embodiment 3 or 4, wherein the amount of the polymer variant sequence in the unselected sample includes a sequence count. Embodiment 6. The method according to any one of Embodiments 3 to 5, wherein the enrichment data includes an enrichment factor for the first polymer variant sequence based on i) the amount of the first polymer variant sequence in the polymer variant sequence in the unselected sample, and ii) the amount of the first polymer variant sequence in the polymer variant sequence in the subset of the sample selected based on the polymer variant function. Embodiment 7. The method according to Embodiment 6, wherein the concentration factor is a functional label of the first polymer variant sequence. Embodiment 8. The method according to any one of Embodiments 1 to 7, wherein the method comprises carrying out multiple generations of (a) and (b), and using the improved polymer variant of (b) to generate the dataset of (a) in subsequent generations of the multiple generations. Embodiment 9. The method according to Embodiment 8, wherein the multiple generations are at least three or four generations. Embodiment 10. The method according to any one of Embodiments 1 to 9, wherein the polymer variant comprises a protein. Embodiment 11. The method according to Embodiment 10, wherein the protein comprises an enzyme. Embodiment 12. The method according to Embodiment 11, wherein the function includes the activity of the enzyme. Embodiment 13. The method according to any one of Embodiments 1 to 12, wherein the improved polymer variant comprises at least 1000 improved polymer variants. Embodiment 14. The method according to any one of Embodiments 1 to 13, wherein the polymer variant sequence includes a nucleic acid sequence encoding the polymer variant. Embodiment 15. The method according to Embodiment 1, further comprising providing a library of cells containing the polymer variant before obtaining the dataset, wherein the polymer variant comprises a plurality of polymer variants. Embodiment 16. The method according to Embodiment 15, further comprising generating the library of cells containing the polymer variant. Embodiment 17. The method according to Embodiment 16, wherein the production provides a site-saturated nucleic acid molecule library (SSL) encoding a protein having each of 20 naturally occurring amino acids at each position. Embodiment 18. The method according to Embodiment 17, further comprising performing gene shuffling and error-prone polymerase chain reaction (epPCR) on the SSL to generate a nucleic acid molecule encoding the polymer variant. Embodiment 19. The method according to Embodiment 18, further comprising transforming cells with the nucleic acid molecule encoding the polymer variant to generate a library of the cells. Embodiment 20. The method according to Embodiment 19, wherein the cells include bacterial cells. Embodiment 21. The method according to Embodiment 15, further comprising distributing the library of cells to generate multiple partitions. Embodiment 22. The method according to Embodiment 21, wherein each of the plurality of partitions contains, on average, one or fewer cells. Embodiment 23. The method according to Embodiment 21, further comprising expanding the cell library in the plurality of partitions. Embodiment 24. The method according to Embodiment 23, further comprising expressing the polymer variant in a library of cells within the plurality of partitions. Embodiment 25. The method of Embodiment 15, further comprising determining the polymer variant function using the library of cells containing the polymer variant. Embodiment 26. The method according to Embodiment 25, wherein the polymer variant comprises a protein. Embodiment 27. The method according to Embodiment 26, wherein the protein comprises an enzyme. Embodiment 28. The method of Embodiment 27, wherein determining the polymer variant function includes using an enzyme substrate and a reaction buffer. Embodiment 29. The method according to Embodiment 28, wherein the polymer variant function includes enzymatic activity. Embodiment 30. The method according to Embodiment 29, wherein the enzyme activity generates a fluorescent product. Embodiment 31. The method of Embodiment 15, further comprising selecting a subset of the cell library containing the polymer variant based on the polymer variant function to provide the sample selected and the unselected sample based on the polymer variant function. Embodiment 32. The method according to Embodiment 31, wherein the function is the amount of enzyme activity. Embodiment 33. The method according to Embodiment 31, wherein the sorting includes sorting the subset of the cell library into a plurality of groups based on the function. Embodiment 34. The method according to Embodiment 32, wherein the plurality of groups is at least four groups. Embodiment 35. The method according to Embodiment 31, wherein a subset of the cell library contains approximately 50 to 100 times the number of unique polymer variants expected to be present in the cell library. Embodiment 36. The method according to Embodiment 31, further comprising determining the polymer variant sequence using the sample selected based on the polymer variant function. Embodiment 37. The method according to Embodiment 36, wherein the sample selected based on the polymer variant function includes a reference sequence. Embodiment 38. The method of Embodiment 36, wherein determining the polymer variant sequence using the sample selected based on the polymer variant function is a method for purifying nucleic acid molecules from the sample selected based on the polymer variant function. Embodiment 39. The method according to Embodiment 31, further comprising purifying nucleic acid molecules from the unselected sample. Embodiment 40. The method according to Embodiment 39, wherein the purification of nucleic acid molecules from the sample selected based on the polymer variant function occurs separately from the purification of nucleic acid molecules from the unselected sample. Embodiment 41. The method according to Embodiment 31, further comprising sequencing the nucleic acid molecules from the sample selected based on the polymer variant function and the nucleic acid molecules from the unselected sample to provide the polymer variant sequence. Embodiment 42. The method according to Embodiment 41, wherein the sequencing is performed by next-generation sequencing. Embodiment 43. The method according to Embodiment 41, further comprising using the polymer variant sequence to determine the amount of the polymer variant sequence in i) the sample selected based on the polymer variant function and ii) the unselected sample. Embodiment 44. The method of Embodiment 43, further comprising determining the enrichment data using i) the abundance of the polymer variant sequence in the sample selected based on the polymer variant function and ii) the abundance of the polymer variant sequence in the unselected sample. Embodiment 45. The method according to Embodiment 44, wherein the concentration data includes a concentration factor. Embodiment 46. The method according to Embodiment 1, further comprising synthesizing the improved polymer variant. Embodiment 47. The method according to Embodiment 46, wherein the synthesis includes synthesis on an oligonucleotide array. Embodiment 48. The method according to Embodiment 47, further comprising assembling the improved polymer variant into a vector. Embodiment 49. The method according to Embodiment 48, wherein the vector is a plasmid. Embodiment 50. The method according to Embodiment 48, further comprising introducing the vector into a host cell. Embodiment 51. The method according to Embodiment 8, wherein the first generation (a) of the multiple generations comprises performing saturated mutagenesis and epPCR to provide a cell library, the second generation (a) of the multiple generations comprises performing epPCR, and the third generation (a) of the multiple generations comprises performing epPCR. Embodiment 52. The method according to Embodiment 51, wherein each of the multiple generations includes multiple sorting steps. Embodiment 53. A non-temporary computer-readable medium that, when executed by a computer processor, includes instructions causing the computer processor to perform the method described in any one of Embodiments 1 to 14. Embodiment 54. A system comprising a computer processor and a non-temporary computer-readable medium as described in Embodiment 53. Embodiment 55. A method for generating improved cells, (a) To generate a library of mutagenic cells, (b) Distributing the library of mutagenic cells into multiple partitions to obtain the distributed mutagenic cells, (c) Screening the distributed mutagenic cells for their phenotype, (d) Based on the screening, select improved cells from the distributed mutagenic cells using a threshold gate, (e) Maintaining the generation frequency of the multiple partitions using the feedback controller (i) of (b), or adjusting the threshold gate using the feedback controller (ii) of (d), (f) A method comprising repeating steps (a) to (d) using the improved cells. Embodiment 56. The method according to Embodiment 55, wherein the method includes maintaining the generation frequency of the plurality of partitions using the feedback controller. Embodiment 57. The method according to Embodiment 56, wherein the plurality of partitions include a plurality of droplets. Embodiment 58. The method according to Embodiment 57, wherein the feedback controller adjusts the flow rate to maintain the frequency of generation of the plurality of droplets. Embodiment 59. The method according to Embodiment 55, wherein the method includes adjusting the threshold gate using the feedback controller. Embodiment 60. The method according to Embodiment 59, wherein the feedback controller raises or lowers the threshold gate. Embodiment 61. The method according to Embodiment 59, wherein the feedback controller adjusts the threshold gate to select a target percentage of the distributed mutagenic cells. Embodiment 62. The method according to Embodiment 55, wherein the feedback controller is a proportional-integral-derivative (PID) algorithm. Embodiment 63. The method according to Embodiment 59, wherein the feedback controller takes into account deviations in mechanical equipment, liquid viscosity, or temperature. Embodiment 64. The method of Embodiment 55, wherein generating the library of mutagenic cells comprises contacting the cells with a mutagen. Embodiment 65. The method according to Embodiment 55, wherein generating the library of mutagenic cells comprises transposon mutagenesis, multiple automated genome engineering (mage), genome shuffling, random recombination, or non-homologous end joining. Embodiment 66. The method of Embodiment 55, wherein generating the library of mutagenic cells includes introducing heterologous genetic elements into cells. Embodiment 67. The method according to Embodiment 55, wherein the heterogeneous element comprises a promoter, a 5' untranslated region, a ribozyme, an RNA stability sequence, a nucleic acid encoding a secretory peptide, a nucleic acid encoding a signal peptide, a nucleic acid encoding a fusion protein, a nucleic acid encoding a DNA binding domain, a nucleic acid encoding a protein-protein binding region, a nucleic acid encoding a codon optimization gene, a 3' untranslated region, or a terminator. Embodiment 68. The method according to Embodiment 55, wherein the mutagenic cells include at least 1,000 cells, at least 10,000 cells, or at least 1,000,000 cells. Embodiment 69. The method of Embodiment 55, further comprising expanding and growing a library of mutagenic cells prior to (b). Embodiment 70. The method according to Embodiment 67, wherein the expansion and growth occur in liquid culture. Embodiment 71. The method according to Embodiment 55, wherein the distribution is controlled by a computer-readable medium. Embodiment 72. The method according to Embodiment 55, wherein each of the plurality of partitions contains a volume of liquid within a predetermined range. Embodiment 73. The method according to Embodiment 55, further comprising imaging each partition of a plurality of partitions prior to (c). Embodiment 74. The method according to Embodiment 71, further comprising ensuring the distribution of the plurality of partitions containing a desired number of mutagenic cells based on the imaging. Embodiment 75. The method according to Embodiment 55, further comprising subjecting the plurality of partitions to cell proliferation conditions between (b) and (c). Embodiment 76. The method according to Embodiment 55, wherein providing the control of the temperature, carbon source, pH, culture medium components, or aeration of the partitions of a subset of the plurality of partitions. Embodiment 77. The method of Embodiment 55, further comprising monitoring cell proliferation in the plurality of partitions subjected to the cell proliferation conditions using imaging. Embodiment 78. The method according to Embodiment 55, wherein the screening includes adding a detection reagent to each of the plurality of partitions. Embodiment 79. The method according to Embodiment 76, wherein the detection reagent changes the environment of each of the multiple partitions. Embodiment 80. The method according to Embodiment 76, wherein the detection reagent comprises a pH adjuster, an ionic strength adjuster, an ion, a substrate, an inhibitor and an enzyme, or a competing agent. Embodiment 81. The method according to Embodiment 55, wherein the library of mutagenic cells encodes the molecule of the choice. Embodiment 82. The method according to Embodiment 79, wherein the screening includes detecting a molecule of interest, measuring the activity of the molecule of interest, wherein the activity is catalytic activity, and the catalytic activity generates a fluorescent product, determining the amount of the molecule of interest, or determining the function of the molecule of interest. Embodiment 83. The method according to Embodiment 55, wherein the screening includes performing a plurality of screenings. Embodiment 84. The method of Embodiment 55, wherein the plurality of screenings include applying a stringent threshold to the target phenotype in subsequent screenings of the plurality of screenings compared to screenings prior to the plurality of screenings. Embodiment 85. The method according to Embodiment 55, wherein selecting the partition includes performing a spectroscopic measurement. Embodiment 86. The method according to Embodiment 55, wherein selecting the partition includes applying a threshold. Embodiment 87. The method of Embodiment 55, wherein (f) comprises removing mutagenic cells from a selected partition to obtain improved cells. Embodiment 88. The method according to Embodiment 87, further comprising maintaining the survival of the mutagenic cells. Embodiment 89. The method according to Embodiment 87, further comprising growing the mutagenic cells. Embodiment 90. The method of Embodiment 55, further comprising screening cells improved for the desired phenotype. Embodiment 91. The method according to Embodiment 55, wherein the mutagenic cells include bacteria. Embodiment 92. A computer-readable medium that, when executed by a computer processor, includes instructions causing the computer processor to perform steps (b) to (e) described in Embodiment 55. Embodiment 93. A system comprising a computer processor and a computer-readable medium as described in Embodiment 92. Embodiment 94. An integrated system for selecting a desired biological variant, comprising two or more integrated units selected from the group consisting of strain optimization units, molecular optimization units, and molecular discovery units, wherein (a) the strain optimization unit comprises a plurality of separate strain optimization reaction vessels, each having a different strain optimization genetic variant or fermentation conditions; (b) the molecular optimization unit comprises a plurality of separate molecular optimization reaction vessels, each having a different target protein variant or application conditions; and (c) the molecular discovery unit comprises a plurality of separate molecular discovery reaction vessels, each having a different discovery molecule or application conditions. Embodiment 95. The integration system according to Embodiment 94, wherein the two or more units are integrated via a computer-readable medium capable of performing one or more of the following functions: (i) storing datasets from each of the two or more integration units in a data repository; (ii) setting threshold parameters for data to be delivered to the data repository; and (iii) giving instructions to each of the two or more units for repeating an input diversity screening. Embodiment 96. The integration system according to Embodiment 94, wherein the two or more units are integrated via a computer-readable medium that includes a machine learning algorithm for integrating data from each of the two or more units with respect to the selected desired biological variant. Embodiment 97. The integrated system according to Embodiment 96, wherein the machine learning algorithm includes an Elastic Net Regression Generalized Linear Model (GLMNET), Support Vector Machine Regression (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Multilayer Perceptron (MLP), or Convolutional Neural Network (CNN). Embodiment 98. The integrated system according to Embodiment 94, wherein each of the strain optimization reaction vessel, the molecule optimization reaction vessel, and / or the molecule discovery reaction vessel contains, on average, one or fewer cells. Embodiment 99. The integrated system according to Embodiment 94, wherein each of the strain optimization reaction vessel, the molecule optimization reaction vessel, and / or the molecule discovery reaction vessel contains, on average, more than one cell. Embodiment 100. The integration system according to Embodiment 99, wherein the average of more than one cell per well is a clonal population. Embodiment 101. The integrated system according to Embodiment 99, wherein the average number of cells per well is greater than one, and the cell population includes a diverse range of cells. Embodiment 102. The integrated system according to Embodiment 94, wherein each of the strain optimization reaction vessel, the molecule optimization reaction vessel, and / or the molecule discovery reaction vessel contains an average of 10 to 100 cells. Embodiment 103. The integrated system according to Embodiment 97, wherein the strain optimization unit, the molecule optimization unit, and the molecule discovery unit are configured to perform single-cell analysis. Embodiment 104. The integrated system according to Embodiment 99, wherein the strain optimization unit, the molecule optimization unit, or the molecule discovery unit is configured to perform multicellular analysis. Embodiment 105. The integrated system according to Embodiment 94, wherein the strain optimization unit, the molecule optimization unit, and / or the molecule discovery unit are configured to screen at least 100 cells per minute. Embodiment 106. The integrated system according to Embodiment 94, wherein each of the two or more integrated units includes a consumption unit. Embodiment 107. The integrated system according to Embodiment 106, wherein the consumption unit includes a microfluidic device. Embodiment 108. The integrated system according to Embodiment 107, wherein the microfluidic device includes one or more functional elements selected from the group including droplet generators, droplet condensers, buffer reservoirs, cell incubators, fluid injectors, droplet splitters, droplet sorters, delay lines, and droplet singulators. Embodiment 109. The integrated system according to Embodiment 94, wherein the reaction vessel of the strain optimization reaction vessel, the molecule optimization reaction vessel, or the molecule discovery reaction vessel has a volume of less than 1 microliter. Embodiment 110. The integrated system according to Embodiment 94, wherein the reaction vessels of the strain optimization reaction vessel, the molecule optimization reaction vessel, and / or the molecule discovery reaction vessel are located within a nanowell array. Embodiment 111. The integrated system according to Embodiment 94, wherein the reaction vessels of the strain optimization reaction vessel, the molecule optimization reaction vessel, and / or the molecule discovery reaction vessel are droplets, plugs, or wells. Embodiment 112. The integrated system according to Embodiment 94, further comprising a detector coupled to one or more of the units, wherein the detector is selected from the group consisting of a mass spectrometer, a Raman spectrometer, a microscope, a cell counter, a fluorescence microscope, an optical microscope, a flow cytometer, a mass spectrometer, a fluorescence plate reader, a near-infrared (NIR) spectrophotometer, and a piezoelectric sensor. Embodiment 113. The integrated system according to Embodiment 94, further comprising a reaction vessel moving actuator selected from the group consisting of a robotic manipulator, a flow generator, an acoustic drop generator, optical tweezers, a thermal drop on demand, and a piezoelectric drop on demand. Embodiment 114. The integration system according to Embodiment 94, wherein each of the different strain-optimized genetic variants and each of the target protein variants are located in different cells, and / or each of the target discovery molecules is located in different cells. Embodiment 115. The integrated system according to Embodiment 94, wherein each strain-optimized genetic variant is generated by random mutagenesis, multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), or CRISPR-guided DNA polymerase. Embodiment 116. The integrated system according to Embodiment 115, wherein the random mutagenesis is saturation mutagenesis. Embodiment 117. The integrated system according to Embodiment 94, wherein each target protein variant is generated by random mutagenesis, multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), or CRISPR-guided DNA polymerase. Embodiment 118. The integrated system according to Embodiment 117, wherein the random mutagenesis is saturation mutagenesis or cyclic substitution. Embodiment 119. The integration system according to Embodiment 94, wherein each discovered molecule is a DNA-coding molecule found in a metagenomic library, a DNA-coding library, a library of DNA from one or more microorganisms, or a library generated by a function generator. Embodiment 120. The integrated system according to Embodiment 94, wherein the strain optimization unit includes a strain optimization library containing multiple different genetic variants of the first strain. Embodiment 121. The integration system according to Embodiment 120, wherein the plurality of different genetic variants are classified by metabolic pathway. Embodiment 122. The integrated system according to Embodiment 121, wherein the strain optimization library includes phenotypic, sequence, and metabolic pathway data. Embodiment 123. The integrated system according to Embodiment 122, wherein the stock optimization unit includes a machine learning algorithm. Embodiment 124. The integrated system according to Embodiment 123, wherein the machine learning algorithm identifies genetic variants optimized across multiple fermentation conditions. Embodiment 125. The integration system according to Embodiment 94, wherein each of the two or more units includes a recursive cycle that includes additional inputs of biological diversity or conditional diversity. Embodiment 126. The plurality of separate reaction volumes are at least 10 5 The integrated system according to Embodiment 94, comprising several separate reaction volumes. Embodiment 127. The integrated system according to Embodiment 94, wherein the application conditions include one or more of pH, substrate, reaction buffer, and temperature. Embodiment 128. The integrated system according to Embodiment 94, wherein the fermentation conditions include one or more of the following: pH, oxygenation, carbon source, buffer concentration, and temperature. Embodiment 129. The integration system according to Embodiment 120, wherein the stock optimization library includes at least 100 different variants of the first stock. Embodiment 130. The integration system according to Embodiment 120, wherein the strain optimization library is generated by random mutagenesis of strains. Embodiment 131. The integrated system according to Embodiment 120, wherein the strain optimization library is selected from the group consisting of a whole mutagenesis library, a genome shuffle library, a target genome library, and a transposon library. Embodiment 132. The integrated system according to Embodiment 131, wherein the whole-genome mutagenesis library is a whole-genome random mutagenesis library. Embodiment 133. The integration system according to Embodiment 131, wherein the target genome library is a promoter swap library. Embodiment 134. The integrated system according to Embodiment 131, wherein the transposon library is a promoter insertion library. Embodiment 135. The integrated system according to Embodiment 94, wherein the molecular optimization unit includes a molecular optimization library containing multiple different genetic variants of the first molecule. Embodiment 136. The integrated system according to Embodiment 135, wherein the molecular optimization library is selected from the group consisting of a single-gene random mutation library, a site-saturated library, a small insert metagenomic library, and a large insert metagenomic library. Embodiment 137. An integrated system for selecting a desired biological variant, comprising a biological diversity data stream and a conditional diversity data stream, wherein (a) each of the data streams is independently programmable, (b) the biological diversity data stream comprises data from screening for different genomic variants of a target protein or a strain expressing the target protein, (c) the conditional diversity data stream comprises data from screening for different conditions of the strain and the target protein, and (d) the biological diversity data stream and the conditional diversity data stream are integrated into an optimization unit configured to receive a target parameter, wherein the optimization unit is instructed to introduce additional biological diversity or conditional diversity until a target protein or target strain matching the target parameter is identified, or ii) the optimization unit identifies a target protein or target strain matching the target parameter. Embodiment 138. The integrated system according to Embodiment 137, wherein the optimization unit includes a computer-readable medium capable of performing one or more of the following functions: (i) storing datasets from the respective biological diversity data streams and the conditional diversity data streams in a data repository; (ii) setting threshold parameters for data to be delivered to the data repository; and (iii) instructing the optimization unit to repeat the input diversity screening. Embodiment 139. The integration system according to Embodiment 137, wherein the optimization unit includes a computer-readable medium containing a machine learning algorithm for integrating the biological diversity data stream and the conditional diversity data stream for the desired selection of biological variants. Embodiment 140. The integrated system according to Embodiment 139, wherein the machine learning algorithm performs Elastic Network Regression Generalized Linear Model (GLMNET), Support Vector Machine Regression (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), Multilayer Perceptron (MLP), or Convolutional Neural Network (CNN). Embodiment 141. The integrated system according to Embodiment 137, wherein the strain optimization unit is configured to screen the target protein or 100 different genomic variants of the strain expressing the target protein per minute. Embodiment 142. The integrated system according to Embodiment 137, wherein the strain optimization unit is configured to screen 100 cells per minute. Embodiment 143. The integrated system according to Embodiment 137, wherein each of the stock optimization units comprises a consumption unit. Embodiment 144. The integrated system according to Embodiment 143, wherein the consumption unit includes a microfluidic device. Embodiment 145. The integrated system according to Embodiment 144, wherein the microfluidic device includes one or more functional elements selected from the group including droplet generators, droplet condensers, buffer reservoirs, cell incubators, fluid injectors, droplet splitters, droplet sorters, delay lines, and droplet singulators. Embodiment 146. The integrated system according to Embodiment 144, wherein the microfluidic device includes one or more reaction vessels. Embodiment 147. The integrated system according to Embodiment 146, wherein the reaction vessel has a volume of less than 1 microliter. Embodiment 148. The integrated system according to Embodiment 146, wherein the reaction vessel is located within a nanowell array. Embodiment 149. The integrated system according to Embodiment 146, wherein the reaction vessel is a droplet, a plug, or a well. Embodiment 150. The integrated system according to Embodiment 149, wherein the droplet is a water-in-oil droplet. Embodiment 151. The integrated system according to Embodiment 149, wherein the droplet is a water droplet on a surface. Embodiment 152. The integration system according to Embodiment 146, wherein the reaction vessels include cells such that there is an average of about one cell in each of two or more reaction vessels. Embodiment 153. The integrated system according to Embodiment 137, further comprising a detector coupled to the optimization unit, wherein the detector is selected from the group consisting of a mass spectrometer, a Raman spectrometer, a microscope, a cell counter, a fluorescence microscope, an optical microscope, a flow cytometer, a mass spectrometer, a fluorescence plate reader, a near-infrared (NIR) spectrophotometer, and a piezoelectric sensor. Embodiment 154. The integrated system according to Embodiment 137, further comprising a reaction vessel moving actuator selected from the group consisting of a robotic manipulator, a flow generator, an acoustic drop generator, optical tweezers, a thermal drop on demand, and a piezoelectric drop on demand. Embodiment 155. The integration system according to Embodiment 155, wherein each of the different genomic variants of the target protein is located in a different cell. Embodiment 156. The integration system according to Embodiment 137, wherein each of the different genomic variants of the strain is generated by random mutagenesis, multiple automated genome engineering (MAGE), or traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), or CRISPR-guided DNA polymerase. Embodiment 157. The integrated system according to Embodiment 156, wherein the random mutagenesis is saturation mutagenesis or cyclic substitution. Embodiment 158. The integration system according to Embodiment 137, wherein each of the different genomic variants of the target protein variant is generated by random mutagenesis, multiple automated genome engineering (MAGE), or traceable genome engineering (CREATE) via CRISPR technology. Embodiment 159. The integrated system according to Embodiment 158, wherein the random mutagenesis is saturated mutagenesis and cyclic substitution. Embodiment 160. A method for selecting a final desired genetic variant, comprising combining data from a diversity screening of biologically diverse genetic elements, including genetic elements of a target protein and genetic elements of a strain, with data from a conditional screening of diverse conditions, including conditions for expression of a target protein and conditions for performance of a target protein, and using the combined dataset to identify the final desired genetic variant. Embodiment 161. The method according to Embodiment 160, wherein the conditional screening is performed on one or more genetic elements that are part of a diversity screening. Embodiment 162. The method according to Embodiment 160, further comprising each cell generating a library of first cells containing different genetic elements from a library of first genetic elements. Embodiment 163. The method of Embodiment 162, further comprising each cell generating a second library of cells containing different genetic elements from a second library of genetic elements, wherein the first library of genetic elements and the second library of genetic elements are screened independently under a plurality of conditions. Embodiment 164. The method according to Embodiment 163, wherein the library of the first genetic element and the library of the second genetic element are screened using a next-generation sequencing (NGS) sequencing system to identify enriched mutations under the plurality of conditions. Embodiment 165. The method according to Embodiment 160, wherein the diversity screening is performed on variants related to strain optimization. Embodiment 166. The method according to Embodiment 165, wherein the strain optimization results in increased yield, improved fermentation conditions, increased viability, and increased tolerance to ion concentration. Embodiment 167. The method according to Embodiment 166, wherein the increase in yield is an increase in protein expression and / or an increase in cell proliferation. Embodiment 168. The method according to Embodiment 160, wherein using the combined dataset to identify the final desired genetic variant further comprises a recursive cycle of (a) identifying one or more intermediate desired genetic variants having the desired trait, (b) obtaining a recombinant library based on the genetic elements of the one or more intermediate desired genetic variants having the desired trait, (c) screening the recombinant library by conditional screening under diverse conditions, and (d) identifying the final desired genetic variant, or (a) identifying one or more intermediate desired genetic variants having the desired trait. Embodiment 169. The method according to Embodiment 160, wherein the diversity screening is performed on variants related to the discovery of a target protein. Embodiment 170. The method according to Embodiment 160, wherein the diversity screening is performed on variants related to target protein optimization. Embodiment 171. The method according to Embodiment 160, wherein each of the biologically diverse genetic elements is screened in parallel. Embodiment 172. The method according to Embodiment 160, wherein each of the biologically diverse genetic elements is screened sequentially. Embodiment 173. The method according to Embodiment 160, wherein the combined dataset includes data from one or more assays of fluorescence generation assays, colorimetric assays, and reporter cell-based assays. Embodiment 174. The method according to Embodiment 173, wherein the fluorescence generation assay is a direct assay, an indirect assay, a binding assay, or an in vivo reporter assay. Embodiment 175. The method according to Embodiment 160, wherein the conditional screening includes adjusting temperature, culture medium, oxygenation, salinity, pH, mixing, reaction time, metal ion concentration, additive concentration, feed rate, or a combination thereof. Embodiment 176. The method of Embodiment 160, wherein the conditional screening comprises performing two or more different conditions in tandem for a first genetic element. Embodiment 177. The method according to Embodiment 160, wherein each of the various conditions is performed in parallel. Embodiment 178. The method according to Embodiment 160, wherein each of the various conditions is executed in series. Embodiment 179. The method of Embodiment 160, further comprising generating a library of diverse genetic elements. Embodiment 180. The method according to Embodiment 179, wherein the generation is carried out by random mutagenesis, site-directed mutagenesis, site saturation, or genome shuffling. Embodiment 181. The method according to Embodiment 179, further comprising introducing each genetic element from the library of diverse genetic elements into cells or a cell-free system. Embodiment 182. The method according to Embodiment 179, further comprising separating cells having genetic diversity from the library of diverse genetic elements into different reaction vessels such that each reaction vessel has one or fewer cells. Embodiment 183. The method according to Embodiment 182, wherein the reaction vessel is a droplet, a plug, or a well. Embodiment 184. The method according to Embodiment 160, wherein the desired genetic variant is related to strain optimization, target molecule optimization, or molecular discovery from an environmental DNA or RNA library. Embodiment 185. The method according to Embodiment 160, wherein the biological diversity is strain-optimized. Embodiment 186. The method of Embodiment 160, wherein identifying the final desired genetic variant comprises measuring one or more properties of the strain, including expression levels, cell viability, cell proliferation, and sensitivity to divalent cation concentration. Embodiment 187. The method according to Embodiment 162, further comprising generating a library of modified strains by site-directed mutagenesis, random mutagenesis, multiple automated genome engineering (MAGE), traceable genome engineering via CRISPR technology (CREATE), transposon-based mutagenesis, genome shuffling, random recombination, non-homologous end joining, traceable multiple recombination (TRMR), or CRISPR-guided DNA polymerase. Embodiment 188. The method according to Embodiment 187, wherein the random mutagenesis is saturation mutagenesis. Embodiment 189. The method according to Embodiment 160, wherein the biological diversity is for target protein optimization. Embodiment 190. The method of Embodiment 189, wherein identifying the desired genetic variant involves measuring one or more properties of the target protein, including substrate specificity, substrate state, sensitivity to divalent cation concentration, temperature tolerance, catalytic performance, resistance to protease deg...
Claims
1. A method comprising: inputting data representative of performance assays and cell culture parameters into a machine learning algorithm, the data being generated by performing performance assays on cells contained within a plurality of partitions having a volume of less than 100 microliters and exposed to cell culture conditions defining the cell culture parameters; selecting a cell as the optimized cell based on the output of the machine learning algorithm; A method comprising:
2. The method of claim 1, wherein the machine learning algorithm outputs data representing proteins.
3. The method described in claim 2, further comprising a step of culturing the optimized cells to produce the protein.
4. The method of claim 2, wherein the protein comprises a structural protein, an enzyme, a surface receptor protein, a peptide hormone, an immune system component, a bioactive peptide, an industrial enzyme, a therapeutic protein, an antibody, an antibody fragment, an antibody conjugate, a polymer-degrading enzyme, a signal transduction factor, a cytokine, and / or a food-based protein.
5. The method of claim 1, wherein the plurality of partitions comprises at least 10,000 partitions.
6. The method of claim 1, wherein the plurality of partitions comprises at least 1,000,000 partitions.
7. The method of claim 1, wherein the performance assay comprises a protein assay.
8. The method of claim 1, wherein the performance assay comprises a cellular assay.
9. The method of claim 1, wherein the performance assay comprises a fluorogenic assay, a chromogenic or colorimetric assay, a reporter cell-based assay, a binding assay and / or an immunological assay.
10. The method of claim 1, wherein the cells contain at least 1,000 different genetic variants.
11. The method of claim 1, wherein the cells contain at least 1,000,000 different genetic variants.
12. The method of claim 1, wherein the cells comprise bacterial cells, yeast, mammalian cells, reptilian cells and / or avian cells.
13. The method of claim 1, wherein the cell culture conditions define cell culture parameters for volumes greater than 100 milliliters.
14. The method of claim 1, wherein the cell culture conditions defining the cell culture parameters include variations in temperature, feed rate, growth or nutrient medium composition, oxygenation, salinity, pH, carbon source, carbon dioxide concentration, buffer concentration, ion concentration, culture period, perfusion or mixing, aeration, reaction time, metal ion concentration, additive concentration, and / or feed schedule, or any one or combination thereof.
15. The method of claim 1, wherein the machine learning algorithm comprises a supervised learning algorithm, an unsupervised learning algorithm, or a reinforcement learning algorithm.
16. The method of claim 1, wherein the machine learning algorithm comprises a linear regression, a logistic regression, a decision tree, a support vector machine, a naive Bayes algorithm, a k-nearest neighbor algorithm, a k-means algorithm, a random forest, a dimensionality reduction algorithm, a gradient boosting algorithm, an XGBoost algorithm, a light gradient boosting algorithm, or a CatBoost algorithm.
17. The method of claim 1, wherein the machine learning algorithm includes a natural language model.
18. The method of claim 1, wherein the machine learning algorithm includes a neural network.
19. A method comprising: inputting data representative of performance assays and cell culture parameters into a machine learning algorithm, the data being generated by performing performance assays on cells contained within a plurality of partitions having a volume of less than 100 microliters and exposed to cell culture conditions defining the cell culture parameters; using the machine learning algorithm to design an optimized protein; A method comprising:
20. The method of claim 19, further comprising a step of synthesizing the optimized protein.
21. The method of claim 20, comprising synthesizing the optimized protein intracellularly.
22. The method of claim 19, wherein the performance assay comprises a protein assay.
23. A method comprising: providing a plurality of partitions, at least some of the partitions containing cells and having a volume of less than 100 microliters; subjecting at least some of the partitions to cell culture conditions defining cell culture parameters; performing a performance assay on said plurality of partitions; inputting data representative of the performance assay and the cell culture parameters into a machine learning algorithm; selecting a cell as the optimized cell based on the output of the machine learning algorithm; A method comprising:
24. The method described in claim 23, further comprising repeating the steps described in claim 23 using the optimized cells as next-generation cells for the multiple partitions.
25. The method described in claim 23, further comprising repeating the steps described in claim 23 at least 10 times using the optimized cells as next-generation cells for the multiple partitions.
26. The method described in claim 23, further comprising repeating the steps described in claim 23 at least 100 times using the optimized cells as next-generation cells for the multiple partitions.
27. A method comprising: providing a plurality of partitions, at least some of the partitions containing cells and having a volume of less than 100 microliters; subjecting at least some of the partitions to cell culture conditions defining cell culture parameters; performing a performance assay on said plurality of partitions; inputting data representative of the performance assay and the cell culture parameters into a machine learning algorithm; using the machine learning algorithm to design an optimized protein; A method comprising:
