Methods and systems for developing mixing protocols

Abstract

A method for developing a predictive model may include identifying a mixture criterion parameter for the predictive model, identifying an evaluation criterion for the predictive model, selecting test values ​​for the mixture criterion parameter, identifying computational fluid dynamics (CFD) simulations to be performed to generate the evaluation criterion, performing the CFD simulation for each combination of test values ​​to generate an evaluation criterion corresponding to each combination of test values, generating a potential predictive model domain that correlates the mixture criterion parameter with the evaluation criterion, identifying a pool of candidate predictive models from the potential predictive model domain, and sorting the pool of candidate predictive models.

Description

Technical Field

[0001] This patent application claims priority to U.S. Provisional Patent Application No. 63 / 166504, filed March 26, 2021, and U.S. Provisional Patent Application No. 63 / 298880, filed January 12, 2022, both of which are incorporated herein by reference in their entirety. Prior Technology

[0002] This disclosure relates to systems and methods for developing and implementing hybrid criteria. Some aspects of this disclosure relate to systems and methods for high-throughput evaluation of hybrid criteria related to the bioproduction of therapeutics.

[0003] Biopharmaceutical products (such as antibodies, fusion proteins, adeno-associated viruses (AAVs), proteins, tissues, cells, peptides, or other biologically derived therapeutics) are increasingly used for the treatment and prevention of infectious diseases, genetic diseases, autoimmune diseases, and other illnesses. The manufacture of biopharmaceutical products requires precise and consistent conditions. To ensure consistency in solutions, including biopharmaceutical products, a blending principle can be employed throughout the manufacturing process. Blending principles help maintain the proper distribution of solution components (such as biopharmaceutical products, cellular waste, host proteins, extracellular nutrients, and other molecules) in the various solutions involved in the manufacture of biopharmaceutical products.

[0004] Mixing criteria can include parameters such as the shape and size of the mixing vessel, the direction and rate of fluid flow within the solution, and the physicochemical properties of the solution. Mixing criteria can be developed for various types of biopharmaceutical products, mixing vessel geometries, culture medium compositions, and host cells. Modifications to biopharmaceutical products, mixing vessel geometries, culture medium compositions, or host cells may require the redevelopment of mixing criteria. Traditional methods for developing mixing criteria are time- and labor-intensive and can lead to substandard mixing criteria. Summary of the Invention

[0005] Embodiments of this disclosure pertain to a method for developing a predictive model. The method may include identifying mixture criterion parameters for the predictive model, identifying evaluation criteria for the predictive model, and / or selecting test values ​​for the mixture criterion parameters. The method may also include identifying the computational fluid dynamics (CFD) simulations required to generate the evaluation criteria. The method may further include performing CFD simulations for each combination of test values ​​to generate evaluation criteria corresponding to each combination of test values. The method may also include generating a potential predictive model domain that associates the mixture criterion parameters with the evaluation criteria, identifying a pool of candidate predictive models from the potential predictive model domain, and / or ranking the pool of candidate predictive models.

[0006] In some embodiments of this disclosure, the mixing criterion parameters may include two or more of the following: impeller velocity, batch size, solution viscosity, solution density, mixing vessel size, and mixing vessel geometry. Evaluation criteria may include two or more of the following: flow pattern, fluid velocity distribution, fluid flow vector field, fluid flow streamlines, steady-state mixing time, transient mixing time, residence time distribution, contour shear strain rate, mean shear strain rate, exposure analysis, and power consumption. The validated CFD simulations may include steady-state flow analysis, transient flow analysis, mixing time analysis, and / or exposure analysis. In some embodiments, the method for developing a mixing prediction model may further include, after generating the potential prediction model domain and before confirming the candidate prediction model pool, calculating the variability inflation factor for each potential prediction model in the potential prediction model domain, and removing potential prediction models from the potential prediction model domain whose variability inflation factor is greater than or equal to a collinearity threshold, thereby generating a subset of potential prediction models. The candidate predictive model pool may include univariate models from a subset whose R² values ​​are higher than all other univariate models in the subset, and bivariate models from a subset whose R² values ​​are higher than all other bivariate models in the subset. Ranking the candidate predictive model pool may include ranking the candidate predictive model pool based on the number of terms, ranking the candidate predictive model pool based on the R² value, or both. In some embodiments of this disclosure, the test value is a first test value, and the method for developing predictive models further includes using candidate predictive models from the candidate predictive model pool to generate an estimate of the evaluation criteria for combinations corresponding to a second test value. Furthermore, the method may further include: performing CFD simulations on combinations of the second test value to generate an evaluation criterion for combinations corresponding to the second test value, and comparing the evaluation criterion for combinations corresponding to the second test value with the estimate of the evaluation criterion for combinations corresponding to the second test value.

[0007] Further embodiments of this disclosure may include a method for developing a predictive model. This method may include identifying first, second, and third mixed criterion parameters for the predictive model; identifying first and second evaluation criteria for the predictive model; selecting a first test value for the first mixed criterion parameter; selecting a second test value for the second mixed criterion parameter; and / or selecting a third test value for the third mixed criterion parameter. The method may also include identifying a first computational fluid dynamics (CFD) simulation to be performed to generate the first evaluation criterion; identifying a second CFD simulation to be performed to generate the second evaluation criterion; generating a first evaluation criterion corresponding to each combination of the first, second, and third test values ​​by performing the first CFD simulation on each combination of the first, second, and third test values; and / or generating a second evaluation criterion corresponding to each combination of the first, second, and third test values ​​by performing the second CFD simulation on each combination of the first, second, and third test values. The method may further include generating a first domain of a first prediction model that associates the first, second, and third mixed criterion parameters with a first evaluation criterion and / or generating a second domain of a second prediction model that associates the first, second, and third mixed criterion parameters with a second evaluation criterion.

[0008] In some embodiments of this disclosure, the method for developing predictive models may further include: calculating the variance inflation factor for each first predictive model and each second predictive model; removing first predictive models with a variance inflation factor greater than or equal to three from a first domain of the first predictive models to generate a first subset of the first predictive models; removing second predictive models with a variance inflation factor greater than or equal to three from a second domain of the second predictive models to generate a second subset of the second predictive models; confirming a first pool of candidate first predictive models, the first pool including univariate models from the first subset whose R² values ​​are higher than all other univariate models in the first subset, bivariate models from the first subset whose R² values ​​are higher than all other bivariate models in the first subset, and trivariate models from the first subset whose R² values ​​are higher than all other trivariate models in the first subset; and confirming a second pool of candidate second predictive models, the second pool including univariate models from the second subset whose R² values ​​are higher than all other univariate models in the second subset, bivariate models from the second subset whose R² values ​​are higher than all other bivariate models in the second subset, and trivariate models from the second subset whose R² values ​​are higher than all other bivariate models in the second subset. The three-variable model whose value is higher than all other three-variable models in the second subset; the fourth test value selected for the first mixture criterion parameter; the fifth test value selected for the second mixture criterion parameter; the sixth test value selected for the third mixture criterion parameter; using each candidate first predictive model in the first pool of candidate first predictive models, generating an estimated first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values; generating the first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values ​​by performing a first CFD simulation on each combination of the fourth, fifth, and sixth test values; and comparing the estimated first evaluation criterion generated by each candidate first predictive model in the first pool of candidate first predictive models with the evaluation criterion corresponding to the fourth, fifth, and sixth test values. The process involves comparing each combination with a first evaluation criterion, using each candidate second prediction model from the second pool of candidate second prediction models to generate an estimated second evaluation criterion for each combination corresponding to the fourth, fifth, and sixth test values, generating a second evaluation criterion for each combination corresponding to the fourth, fifth, and sixth test values ​​through a second CFD simulation, comparing the estimated second evaluation criterion generated by each candidate second prediction model from the second pool of candidate second prediction models with the second evaluation criterion for each combination corresponding to the fourth, fifth, and sixth test values, and comparing the estimated first evaluation criterion with the first evaluation criterion for each combination corresponding to the fourth, fifth, and sixth test values.A first prediction model is selected from the first pool of candidate first prediction models; a second prediction model is selected from the second pool of candidate second prediction models based on a comparison of the estimated first evaluation criterion with the first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values; the first prediction model is used to determine the first evaluation criterion corresponding to the mixed criterion; and the second prediction model is used to determine the second evaluation criterion corresponding to the mixed criterion.

[0009] Further embodiments of this disclosure may include a method for modeling shear strain associated with a hybrid criterion. This method may include identifying hybrid criterion parameters for a prediction model, selecting test values ​​for the hybrid criterion parameters, performing computational fluid dynamics exposure analysis on each combination of test values ​​to generate shear strain corresponding to each combination of test values, identifying a pool of candidate prediction models, ranking the pool of candidate prediction models, selecting a prediction model from the pool of candidate prediction models, and using the prediction model to evaluate the cumulative shear strain of the hybrid criterion at multiple time intervals to generate shear strain histogram data.

[0010] In some embodiments of this disclosure, a method for modeling shear strain associated with a mixing criterion includes wherein the mixing criterion parameters include two or more of the following: impeller velocity, batch size, solution viscosity, solution density, mixing vessel size, and mixing vessel geometry. Ranking the pool of candidate predictive models includes ranking the pool based on the number of items, ranking the pool based on the R² value, or both, and selecting a predictive model from the pool includes selecting the model with the highest R² value. The mixing criterion may be a mixing criterion associated with a biopharmaceutical product in a bioreactor. The method may further include using shear strain histogram data to assess the risk of visible or sub-visible particle formation. Simple Explanation of the Diagram

[0011] The drawings included in and forming part of this specification illustrate various exemplary embodiments and, together with the description herein, serve to explain the principles of the disclosed embodiments. Any feature of the embodiments or examples described herein (e.g., compositions, formulations, methods, etc.) may be combined with any other embodiments or examples, and all such combinations are included in this disclosure. Furthermore, the described systems and methods are neither limited to any single aspect or embodiment thereof, nor to any combination or arrangement of such aspects and embodiments. For the sake of brevity, certain arrangements and combinations are not discussed and / or described separately herein.

[0012] Figure 1 depicts a strain histogram according to an aspect of this disclosure.

[0013] Figure 2 illustrates, in flowchart form, an exemplary method for developing a predictive model for evaluating hybrid criteria according to aspects of this disclosure.

[0014] Figures 3A and 3B are graphical representations of a mixing container according to aspects of this disclosure.

[0015] Figure 4A is a visual depiction of a fluid flow vector field according to an aspect of this disclosure.

[0016] Figure 4B is a visual depiction of fluid flow streamlines according to an aspect of this disclosure.

[0017] Figure 4C is a visual depiction of the contour shear strain rate according to an aspect of this disclosure.

[0018] Figure 5 illustrates, in flowchart form, an exemplary method for constructing a potential predictive model according to an aspect of this disclosure.

[0019] Figure 6 illustrates the relationship between the mixing time determined by CFD analysis and the mixing time determined by the prediction model according to aspects of this disclosure.

[0020] Figure 7 illustrates the relationship between strain rates determined by CFD analysis and strain rates determined by prediction models according to aspects of this disclosure.

[0021] Figure 8 depicts a strain rate histogram generated by plotting a prediction model according to aspects of this disclosure.

[0022] Figures 9A to 9C are visual depictions of the theoretical mechanism of aggregate formation according to aspects of this disclosure.

[0023] Figure 10A is a visual depiction of the vertical velocity contour lines according to an aspect of this disclosure.

[0024] Figure 10B is a visual depiction of the volume-average velocity according to an aspect of this disclosure.

[0025] Figure 11 depicts the vertical velocity as a function of the tank radius according to one aspect of this disclosure. Implementation

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. While any suitable methods and materials (e.g., similar to or equivalent to those described herein) may be used in the practice or testing of this disclosure, specific example methods are described herein. All publications mentioned are incorporated herein by reference.

[0027] As used herein, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements may include not only those elements but also other elements not expressly listed or inherent to such process, method, article, or apparatus. The term "exemplary" is used in the sense of "example" rather than "ideal." The terms "for example" and "such" and their grammatically equivalent words are to be understood as following the usage of "and not limited to," unless otherwise expressly stated.

[0028] As used herein, the term "about" is intended to describe the variation caused by experimental error. When applied to numerical values, the term "about" may mean a difference of + / - 5% from a published value, unless a different variation has been specified. As used herein, the singular forms "a" and "the" include the plural referent, unless the context clearly specifies otherwise. Furthermore, all ranges should be understood to include endpoints; for example, 1 centimeter (cm) to 5 centimeters would include lengths of 1 cm and 5 cm, and all distances between 1 cm and 5 cm.

[0029] It should be noted that, unless otherwise specified, all values ​​disclosed or claimed herein (including all disclosed values, limitations and ranges) may vary by + / - 5% from the disclosed values.

[0030] As used herein, the term "peptide" refers to any amino acid polymer having more than about 20 amino acids covalently linked by amide bonds. Proteins comprise one or more chains of amino acid polymers (e.g., peptides). Thus, a peptide can be a protein, and a protein can comprise multiple peptides to form a single functional biomolecule.

[0031] Post-translational modifications can alter or change the structure of a polypeptide. For example, disulfide bonds (such as SS bonds between cysteine ​​residues) can form after translation in some proteins. Some disulfide bonds are essential for the correct structure, function, and interactions of polypeptides, immunoglobulins, proteins, cofactors, and matrix components. Besides disulfide bond formation, proteins can undergo other post-translational modifications, such as lipoylation (e.g., myristoylation, palmitoylation, farnesoylation, geranylgeranylation, and glycosylphosphatidylinositol (GPI) anchor formation), alkylation (e.g., methylation), acetylation, acetylation, glycosylation (e.g., adding glycosyl groups to arginine, aspartic acid, cysteine, hydroxylysine, serine, threonine, tyrosine, and / or tryptophan), and phosphorylation (i.e., adding phosphate groups to serine, threonine, tyrosine, and / or histidine). Post-translational modifications can affect hydrophobicity, electrostatic surface properties, or other peptide-dependent properties that determine surface interactions.

[0032] As used herein, the term "protein" includes biotherapeutic proteins, recombinant proteins used for research or treatment, trap proteins and other Fc fusion proteins, chimeric proteins, antibodies, monoclonal antibodies, human antibodies, bispecific antibodies, antibody fragments, antibody-like molecules, nanoantibodies, recombinant antibody chimeras, cytokines, chemokines, peptide hormones, etc. A protein of interest (POI) may include any polypeptide or protein that is to be isolated, purified, or otherwise prepared. POIs may include cellularly produced polypeptides, including antibodies.

[0033] As used herein, the term "antibody" refers to an immunoglobulin composed of four polypeptide chains: two heavy (H) chains and two light (L) chains linked together by disulfide bonds. Typically, antibodies have a molecular weight exceeding 100 kDa, such as between 130 kDa and 200 kDa, for example, approximately 140 kDa, 145 kDa, 150 kDa, 155 kDa, or 160 kDa. Each heavy chain contains a heavy chain variable region (HCVR or VH) and a heavy chain constant region. The heavy chain constant region contains three domains: CH1, CH2, and CH3. Each light chain contains one light chain variable region (LCVR or VL) and one light chain constant region. The light chain constant region contains one domain, CL. The VH and VL regions can be further subdivided into highly variable regions called complementarity determining regions (CDRs) and more conserved regions called framework regions (FRs). Each VH and VL consists of 3 CDRs and 4 FRs, arranged in the following order from the amino terminus to the carboxyl terminus: FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4 (heavy chain CDRs can be abbreviated as HCDR1, HCDR2, and HCDR3; light chain CDRs can be abbreviated as LCDR1, LCDR2, and LCDR3).

[0034] One type of immunoglobulin is called immunoglobulin G (IgG), which is common in human serum and consists of four polypeptide chains—two light chains and two heavy chains. Each light chain is linked to one heavy chain via a cystine disulfide bond, and the two heavy chains are linked together by two cystine disulfide bonds. Other types of human immunoglobulins include IgA, IgM, IgD, and IgE. Regarding IgG, there are four subclasses: IgG1, IgG2, IgG3, and IgG4. Each subclass has a different constant region and therefore may have different effector functions. In some embodiments described herein, the POI may include a target polypeptide, including IgG. In at least one embodiment, the target polypeptide includes IgG4.

[0035] As used herein, the term "antibody" also includes the antigen-binding fragment of a complete antibody molecule. As used herein, the terms "antigen-binding portion" and "antigen-binding fragment" of an antibody include any naturally occurring, enzymatically obtained, synthetic, or genetically engineered polypeptide or glycoprotein that specifically binds to an antigen to form a complex. The antigen-binding fragment of an antibody can be derived from, for example, a complete antibody molecule, using any suitable standard technique such as proteolytic digestion or recombinant genetic engineering, including manipulating and expressing DNA encoding variable and optionally constant domains of the antibody. This DNA is known and / or directly available from, for example, commercial sources, DNA libraries (including, for example, phage-antibody libraries), or can be synthesized. DNA can be sequenced or chemically manipulated using molecular biotechnologies, such as arranging one or more variable and / or constant domains into suitable conformations, or introducing codons, generating cysteine ​​residues, modifying, adding, or deleting amino acids, etc.

[0036] Recombinant cell-based production systems can be used to produce target molecules (e.g., target peptides / antibodies), such as insect baculovirus systems, yeast systems (e.g., Pichia sp.), or mammalian systems (e.g., CHO cells and CHO derivatives such as CHO-K1 cells). The term "cell" includes any cell suitable for expressing recombinant nucleic acid sequences. Cells include prokaryotes and eukaryotes (single-celled or multi-celled), bacterial cells (e.g., strains of Escherichia coli, Bacillus, Streptomyces, etc.), mycobacterial cells, fungal cells, yeast cells (e.g., Saccharomyces cerevisiae, Saccharomyces pombe, P. pastoris, P. methylica, etc.), plant cells, insect cells (e.g., SF-9, SF-21, bacculovirus-infected insect cells, Trichoplusiani, etc.), non-human animal cells, human cells, or cell fusions such as fusion tumors or quadromas. In some implementations, the cells may be human, monkey, ape, hamster, rat, or mouse cells. In some embodiments, the cells may be eukaryotic cells and may be selected from the following cell lines: CHO (e.g., CHO K1, DXB-11 CHO, Veggie-CHO), COS (e.g., COS-7), retinal cells, Vero, CV1, kidney cells (e.g., HEK293, 293 EBNA, MSR 293, MDCK, HaK, BHK), HeLa, HepG2, WI38, MRC 5, Colo205, HB 8065, HL-60 (e.g., BHK21), Jurkat, Daudi, A431 (epidermal), CV-1, U937, 3T3, L cells, C127 cells, SP2 / 0, NS-0, MMT 060562, Settly cells, BRL 3A cells, HT1080 cells, myeloma cells, tumor cells, and cell lines derived from the aforementioned cells. In some implementations, the cell may contain one or more viral genes, such as retinal cells expressing viral genes (e.g., PER.C6™ cells).

[0037] The term "target molecule" may be used herein to refer to a target peptide (e.g., an antibody, antibody fragment, or other protein or protein fragment), or to a substance intended to be manufactured, isolated, purified, and / or included in a pharmaceutical product (e.g., adeno-associated virus (AAV) or other molecules for therapeutic purposes). While the methods according to this disclosure may be applicable to target peptides, they may be applicable to other target molecules. For example, AAV may be prepared according to suitable methods (e.g., depth filtration, affinity chromatography, etc.), and mixtures comprising AAV may be subjected to the methods according to this disclosure. Additional procedures (e.g., removal of "empty cassettes" or AAV that does not contain the target sequence) may be performed on mixtures comprising AAV before or after following one or more methods of this disclosure.

[0038] In some embodiments, the target molecule is an antibody, human antibody, humanized antibody, chimeric antibody, monoclonal antibody, multispecific antibody, bispecific antibody, antigen-binding antibody fragment, single-chain antibody, diabody, triabody, or tetrabody, Fab fragment or F(ab')2 fragment, IgD antibody, IgE antibody, IgM antibody, IgG antibody, IgG1 antibody, IgG2 antibody, IgG3 antibody, or IgG4 antibody. In one embodiment, the antibody is an IgG1 antibody. In one embodiment, the antibody is an IgG2 antibody. In one embodiment, the antibody is an IgG4 antibody. In one embodiment, the antibody is a chimeric IgG2 / IgG4 antibody. In one embodiment, the antibody is a chimeric IgG2 / IgG1 antibody. In one embodiment, the antibody is a chimeric IgG2 / IgG1 / IgG4 antibody.

[0039] In some embodiments, the target molecule (e.g., the antibody) is selected from anti-programmed cell death 1 antibody (e.g., the anti-PD1 antibody described in US Patent Application Publication No. US2015 / 0203579A1), anti-programmed cell death ligand-1 (e.g., the anti-PD-L1 antibody described in US Patent Application Publication No. US2015 / 0203580A1), anti-Dll4 antibody, anti-angiopoetin-2 antibody (e.g., the anti-ANG2 antibody described in US Patent No. 9402898), anti-angiopoetin-like 3 antibody (e.g., the anti-AngPtl3 antibody described in US Patent No. 9018356), and anti-platelet-derived growth factor receptor antibody. Antibodies (e.g., the anti-PDGFR antibody described in US Patent No. 9265827), anti-prolactin receptor antibodies (e.g., the anti-PRLR antibody described in US Patent No. 9302015), anti-complement 5 antibodies (e.g., the anti-C5 antibody described in US Patent Application Publication No. US2015 / 0313194A1), anti-TNF antibodies, anti-epidermal growth factor receptor antibodies (e.g., the anti-EGFR antibody described in US Patent No. 9132192 or the anti-EGFRvIII antibody described in US Patent Application Publication No. US2015 / 0259423A1), and anti-proprotein convertase subtilisin Kexin-9 antibodies. Antibodies (such as the anti-PCSK9 antibody described in US Patent No. 8062640 or US Patent Application Publication No. US2014 / 0044730A1), anti-Growth and Differentiation Factor-8 antibodies.Antibodies (e.g., anti-GDF8 antibody as described in US Patent Nos. 8871209 or 9260515, also known as anti-myostatin antibody), anti-glucagon receptor (e.g., anti-GCGR antibody as described in US Patent Application Publication Nos. US2015 / 0337045A1 or US2016 / 0075778A1), anti-VEGF antibody, anti-IL1R antibody, interleukin 4 receptor antibody (e.g., anti-IL4R antibody as described in US Patent Application Publication Nos. US2014 / 0271681A1 or US Patent Nos. 8735095 or 8945559), anti-interleukin 6 receptor antibody (e.g., anti-IL6R antibody as described in US Patent Nos. 7582298, 8043617 or 9173880), anti-interleukin 33 (anti-interleukin... 33)(e.g., anti-IL33 antibody as described in US Patent Application Publication No. US20140271658A1 or US2014 / 0271642A1), anti-respiratory syncytial virus antibody (e.g., anti-RSV antibody as described in US Patent Application No. US2014 / 0271653A1), anti-cluster of differentiation 3 (e.g., anti-CD3 antibody as described in US Patent Application Publication Nos. US2014 / 0088295A1, US20150266966A1, and US Application No. 62 / 222605), anti-cluster of differentiation 20 (e.g., anti-CD20 antibody as described in US Patent Application Publication Nos. US2014 / 0088295A1, US20150266966A1, and US Patent No. 7879984), anti-cluster of differentiation 48 (… Differentiation-48 (e.g., the anti-CD48 antibody described in US Patent No. 9228014), anti-Fel d1 antibody (as described in US Patent No. 9079948), and anti-Middle East Respiratory Syndrome virus (MTRSV) antibody.Anti-virus antibodies (e.g., anti-MERS antibodies), anti-Ebola virus antibodies (e.g., Regeneron's REGN-EB3), anti-CD19 antibodies, anti-CD28 antibodies, anti-IL1 antibodies, anti-IL2 antibodies, anti-IL3 antibodies, anti-IL4 antibodies, anti-IL5 antibodies, anti-IL6 antibodies, anti-IL7 antibodies, anti-Erb3 antibodies, anti-Zika virus antibodies, anti-Lymphocyte Activation Gene 3 antibodies (e.g., anti-LAG3 antibodies or anti-CD223 antibodies), and anti-Activin A antibodies. All U.S. patents and U.S. patent publications mentioned in this paragraph are incorporated herein by reference in their entirety.

[0040] In some embodiments, the target molecule (e.g., a bispecific antibody) is selected from anti-CD3 X anti-CD20 bispecific antibody, anti-CD3 X anti-mucin 16 bispecific antibody, and anti-CD3 X anti-prostate-specific membrane antigen bispecific antibody. In some embodiments, the target molecule is selected from the group consisting of alirocumab, sarilumab, fasinumab, nesvacumab, dupilumab, trevogrumab, evinacumab, and rinucumab.

[0041] In some embodiments, the target molecule is a recombinant protein (e.g., an Fc fusion protein) comprising an Fc moiety and another domain. In some embodiments, the Fc fusion protein is a receptor Fc fusion protein comprising one or more extracellular domains of a receptor coupled to the Fc moiety. In some embodiments, the Fc moiety comprises a hinge region followed by the CH2 and CH3 domains of IgG. In some embodiments, the receptor Fc fusion protein contains two or more distinct receptor chains that bind a single ligand or multiple ligands. For example, the Fc fusion protein is a TRAP protein, such as an IL-1 trap (e.g., rilonacept, which includes an IL-1RAcP ligand-binding region fused to the extracellular region of the Il-1R1 of the Fc fused to hIgG1; see U.S. Patent No. 6,927,004, the entire contents of which are incorporated herein by reference), or a VEGF trap (e.g., aflibercept or ziv-aflibercept, which includes an Ig domain 2 of the VEGF receptor Flk1 fused to the Ig domain 3 of the VEGF receptor Flt1 of the Fc fused to hIgG1; see U.S. Patent Nos. 7,087,411 and 7,279,159, both of which are incorporated herein by reference in their entirety). In other embodiments, the Fc fusion protein is a ScFv-Fc-fusion protein, which includes one or more of one or more antigen-binding domains, such as a variable heavy chain fragment and a variable light chain fragment of an antibody partially coupled to the Fc.

[0042] The term "culture medium" or "culture medium" refers to a nutrient solution used to culture cells, typically providing the nutrients necessary for cell growth, such as carbohydrate energy sources, essential amino acids, trace elements, and vitamins. Culture media may contain extracts, such as serum or protein hydrolysate, which provide the raw materials supporting cell growth. In some embodiments, the culture medium may contain yeast-derived or soybean extracts instead of animal-derived extracts. A chemically defined culture medium is one in which all chemical components are known. A chemically defined culture medium may be completely free of animal-derived components, such as serum or animal-derived protein hydrolysate. The culture medium may also be protein-free. "Fresh culture medium" can refer to a culture medium that has not yet been introduced into a cell culture and / or utilized by the cells in the cell culture. Fresh culture medium may typically contain a high nutrient level and contain very little waste. "Used culture medium" can refer to a culture medium that has been used by cells in cell culture and, compared to fresh culture medium, typically includes a lower nutrient level and a higher water level.

[0043] Typically, mixing guidelines can be incorporated into several stages of biopharmaceutical product manufacturing. For example, during host cell culture or the acquisition of a biopharmaceutical product, mixing guidelines can be used to ensure the proper distribution of the resulting biopharmaceutical product, cells, nutrients, waste, and other components of the culture medium. Mixing guidelines can be used with a container configured to perform the mixing guidelines, also known as a mixing vessel. In some embodiments, a bioreactor can be used as a mixing vessel. In other embodiments, the culture medium can be transferred from the bioreactor to different types of mixing vessels prior to performing the mixing guidelines.

[0044] After obtaining a biopharmaceutical product (e.g., a protein of interest), the obtained product can be kept in solution. Solutions containing biopharmaceutical products may undergo one or more chromatography, filtration (e.g., ultrafiltration, diafiltration, or a combination thereof), or purification (e.g., virus deactivation) steps to improve the purity and efficacy of the biopharmaceutical product. At all stages, mixing principles can be employed to homogenize the solution and / or ensure an appropriate distribution of solution components. In addition to the applications discussed above, mixing principles can be used to combine and / or dilute individual biocontainers, batches, or lots.

[0045] Furthermore, mixing guidelines can be applied to solutions that do not contain the protein of interest. For example, the manufacturing steps of the aforementioned biopharmaceutical products require the use of buffers, culture media, and other solutions. The preparation of buffers, culture media, and other solutions may include the use of one or more mixing guidelines.

[0046] Specific properties of biopharmaceutical products or their manufacturing processes that depend on mixing criteria can be monitored to assess the impact of mixing criterion parameters on the resulting biopharmaceutical products. For example, flow patterns, fluid velocity distributions, fluid flow vector fields, fluid flow streamlines, mixing times (e.g., steady-state or transient mixing times), residence time distributions, contour shear strain rates, mean shear strain rates, exposure analysis, and / or power consumption associated with mixing criteria can be used to evaluate the utility and / or effectiveness of mixing criteria.

[0047] Mixing criteria may include operating parameters for the mixing vessel, such as the vessel's dimensions, impeller speed, load size as a percentage of total capacity, solution viscosity, and / or other operating parameters describing the requirements of the mixing criteria. In some embodiments, the mixing criteria are completed when the solution (including, for example, culture medium, cells, proteins of interest, and / or other molecules) is sufficiently homogenized. The duration of the mixing criteria, i.e., the time required for the solution to achieve sufficient homogeneity, is called the mixing time. The degree of solution mixing can be quantified by a mixing index. The mixing index can be defined as the ratio of the standard deviation of the concentration (e.g., of the protein of interest or other molecules) to the final concentration. Mixing time can be quantified as the amount of time required to achieve a mixing index of approximately 5% under a given mixing criterion.

[0048] In the traditional development of mixing criteria, the physicochemical properties of the protein of interest and the culture medium containing the protein of interest are considered to generate potential mixing criteria. Potential mixing criteria are tested through alternative mixing studies to map the operating range and collect mixing time data. Based on the mixing time data collected from various points within the operating range, one or more candidate mixing criteria can be identified. Candidate mixing criteria can be further tested with shear stress and overmixing studies. Shear stress and overmixing studies generate product quality data that can be used to evaluate candidate mixing criteria.

[0049] Shear stress and overmixing studies must be conducted after mixing time data are generated, as shear stress and overmixing depend on mixing time. If product quality data from shear stress and overmixing studies indicate that the mixing criteria are inappropriate, the development of mixing criteria must be restarted to generate potential mixing criteria. Furthermore, alternative mixing studies must be conducted on the new potential mixing criteria to generate mixing time data that can be used for further shear stress and overmixing studies.

[0050] The traditional development process for hybrid criteria is limited because alternative hybrid studies must be performed to evaluate hybrid criteria that may ultimately lead to unfavorable product quality data. The requirement to run multiple experiments in the traditional development process to determine whether a potential hybrid criterion should be studied results in time- and labor-intensive development of hybrid criteria. Furthermore, events related to the implemented hybrid criteria that could affect the quality of the resulting biopharmaceutical product, such as gas-liquid interface stress, air entrainment, and the risk of visible or subvisible particle formation, are not addressed in the traditional development process.

[0051] In addition to the failure of conventional blending guideline development processes to address all factors that could adversely affect the resulting biopharmaceutical product, scaled studies can lead to excessively high shear stresses. Figure 1 shows a strain histogram illustrating how scaled shear stress studies associated with conventional blending guideline development overestimate shear stress. Curve 610 shows a strain histogram of a scaled shear stress study compared to region 605 of the manufacturing conditions for a validated blending guideline. In other words, region 605 represents the actual shear stress in a validated pharmaceutical blending guideline, while curve 610 represents the predicted shear stress from a scaled study. The illustration in Figure 1 shows that scaled studies exhibit higher shear stresses compared to the typical operating range of a blending guideline.

[0052] Alternative mixing studies, shear stress studies, and overmixing studies associated with the development of traditional mixing criteria do not quantify the risks of gas-liquid interfacial stress, entrained air, and visible or subvisible particle formation. Therefore, these indicators have traditionally been assessed through comprehensive surveys using actual biopharmaceutical products. Conducting comprehensive surveys using products is both expensive and time-consuming. The cost and time constraints of comprehensive surveys reduce reproducibility and increase the difficulty of collecting sufficient samples to reduce sampling variability. Furthermore, probes associated with comprehensive surveys may affect flow rates related to mixing criteria and provide inaccurate data. Because the nature of comprehensive surveys is specific to the parameters of a given mixing criterion, comprehensive surveys often require revalidation.

[0053] The systems and methods disclosed herein provide a development process for improving mixing criteria. For example, the systems and methods described herein can allow the development of predictive models capable of high-throughput evaluation of mixing criteria. Predictive models can be generated that quantify the risks associated with gas-liquid interfacial stress, entrained air, and visible or subvisible particle formation related to mixing criteria.

[0054] Referring to Figure 2, the method 200 for developing a predictive model for evaluating hybrid criteria may include mapping the design space 201, designing a design of experiment (DOE) 202, performing computational fluid dynamics (CFD) analysis 203, constructing candidate predictive models 204, and / or evaluating the predictions 205.

[0055] The mapping design space 201 may include identifying the mixing criterion parameters to be studied. Mixing criterion parameters may include "input variables" or aspects of the mixing criterion that can be adjusted, changed, controlled, and / or monitored to affect the results of the mixing criterion. Examples of mixing criterion parameters include, but are not limited to, impeller speed, batch size, solution viscosity, solution density, mixing vessel size, mixing vessel geometry, and mixing time.

[0056] Impeller speed can be quantified as revolutions per minute (RPM) or as a percentage of maximum impeller speed. Batch size can refer to the volume loaded in the mixing container as a percentage of the mixing container's capacity. Solution viscosity and solution density are parameters specific to the protein of interest. During production, solution viscosity and density can be adjusted to achieve the desired viscosity and density parameters before performing mixing criteria.

[0057] In addition to the aforementioned potential mixing criteria parameters, mapping the design space may include identifying potential mixing container dimensions and potential mixing container geometries. Mixing containers can have various shapes and sizes. For example, mixing containers may include cylindrical, conical, elliptical, square, or combinations thereof. Examples of mixing container geometries are shown in Figures 3A and 3B. The mixing container 100 shown in Figure 3A includes a height and a width, where the height is greater than the width. The mixing container 100 shown in Figure 3B includes a height and a width, where the width is greater than the height. The height-to-width ratio of the mixing container 100 is a key element of the mixing container geometry and can influence the flow pattern of the fluid within the mixing container 100.

[0058] The mixing container 100 may include one or more mechanisms capable of providing agitation. For example, the mixing container 100 may include one or more impellers 110 capable of providing flow within the mixing container. The mixing container 100 shown in FIG3A includes one impeller 110 disposed on one side of the mixing container 100. The mixing container 100 shown in FIG3B includes two impellers 110, symmetrically disposed on opposite sides of the mixing container 100. Alternatively, agitation within the mixing container 100 may be provided by concentrically mounted impellers, wave bags, rocking actuators, or other devices for agitating the solution within the mixing container 100.

[0059] Although exemplary mixing container geometries are shown in Figures 3A and 3B, they are merely two examples. In some embodiments, mixing container 100 may include baffles or other structures designed to alter fluid flow within mixing container 100. Mixing container geometries, including other proportions, configurations, shapes, and mechanisms for providing agitation, may be used with the systems and methods described herein.

[0060] In addition to confirming the mixing criterion parameters, the mapped design space 201 may also include confirming evaluation criteria. Evaluation criteria may include "output variables" or aspects of the mixing criteria, depending on the values ​​selected for the mixing criterion parameters. Examples of evaluation criteria include, but are not limited to, flow pattern, fluid velocity distribution, fluid flow vector field, fluid flow streamlines, mixing time (e.g., steady-state mixing time or transient mixing time), residence time distribution, contour shear strain rate, mean shear strain rate, exposure analysis, power consumption, pressure, turbulent dissipation rate, and Kolmogorov length.

[0061] Referring again to Figure 2, the method 200 for developing a predictive model for evaluating mixed criteria may include constructing a Design of Experiments (DOE) 202. For example, a DOE design may be constructed after the mixed criterion parameters and evaluation criteria have been determined. A Design of Experiments (DOE) is a method for constructing experiments, simulations, and / or measurements that can confirm multivariate interactions. DOEs are as understood by those skilled in the art to which this invention pertains and will not be described in detail further.

[0062] In the context of developing a predictive model for evaluating hybrid criteria, constructing a DOE design 202 involves selecting test values ​​for each validated hybrid criterion parameter and identifying the experiments, simulations, and measurements that must be performed to determine the evaluation criteria for the test values ​​of each set of hybrid criterion parameters.

[0063] For example, if impeller speed, batch size, solution viscosity, and mixing vessel size are identified as four mixing criterion parameters, then constructing a DOE design 202 involves selecting test values ​​for impeller speed, batch size, solution viscosity, and mixing vessel size. In some implementations, approximately 10 to approximately 500 test values ​​may be selected, for example, approximately 30 to approximately 100 test values ​​for each mixing criterion parameter. Other numbers may be selected, such as less than approximately 10, or approximately 100 to approximately 1000 test values ​​for each mixing criterion parameter. The accuracy of subsequent CFD analysis is related to the number of test values ​​selected for each mixing criterion parameter, and selecting more test values ​​for certain mixing criterion parameters can provide more meaningful CFD analysis.

[0064] Referring again to Figure 2, the method 200 for developing a predictive model for evaluating the hybrid criteria may include performing computational fluid dynamics (CFD) analysis. For example, the CFD analysis may be performed after test values ​​for each validated hybrid criterion parameter, and confirm the experiments, simulations, and measurements that must be performed to determine the evaluation criteria for each set of hybrid criterion parameter test values.

[0065] CFD analysis may include one or more simulations indicating fluid flow within the mixing container 100. For example, CFD analysis may include steady-state flow analysis, transient flow analysis, mixing time analysis, and / or exposure analysis. In particular, transient flow analysis can help assess the acceleration time from rest to steady-state velocity, assess the likelihood of bubbling, foaming, or sloshing, and quantify surface deformation (e.g., as part of an aggregate formation risk assessment).

[0066] CFD analysis can be based on mathematical solutions to fluid flow models, including but not limited to conservation laws, Navier-Stokes equations, Euler equations, Bernoulli equations, compression wave equations, boundary layer equations, idealized flow, potential flow, duct flow, eddy formation, vortex formation, and turbulence formation. CFD analysis can be performed on a computer system running CFD analysis software such as Star CCM, OpenFoam, Simulia, and Ansys Workbench.

[0067] In the context of this disclosure, to determine how the geometry of a mixing vessel affects evaluation criteria, one or more mixing vessel geometries can be programmed into a computer system that operates the analysis software. For example, the dimensions and shape of the mixing vessel 100, as well as the dimensions, shape, and arrangement of the mechanisms used to induce agitation (e.g., impeller 110), can be modeled to construct the various flow simulations described above.

[0068] The results of CFD analysis may include vector plots, streamline plots, strain rate contour plots, strain histograms, flow patterns, fluid velocity distribution, fluid flow vector field, fluid flow streamlines, steady-state mixing time, transient mixing time, residence time distribution, contour shear strain rate, mean shear strain rate, exposure analysis, power consumption, pressure, turbulence dissipation rate, and / or Komogorov length.

[0069] Figure 4A shows an exemplary vector plot generated as a result of CFD analysis. The vector plot includes multiple vectors 310. The direction of each vector 310 indicates the direction of fluid flow at the vector location, and the magnitude of the vector indicates the velocity of the fluid flow at the vector location. Figure 4B shows a streamline plot generated as a result of CFD analysis. The streamline plot includes multiple streamlines 320. Each streamline represents a curve tangent to the velocity vector of the flow and indicates where the fluid element will travel in steady state.

[0070] Vector plots and streamline plots can be used to identify stagnant regions, eddies, or other flow structures that can affect the effectiveness of mixing criteria. Vector plots, streamline plots, or both can be used to qualitatively compare different mixing criterion parameters (e.g., different mixing container geometries).

[0071] Figure 4C shows a strain rate contour map depicted in black and white. In practice, different regions of the strain rate contour map can be represented by different colors. Table 1 shows the approximate strain rates associated with marked regions 331 to 336, and exemplary colors that can be used to represent the grids of strain rates shown in Figure 4C and Table 1. [surface] [1] [-with diagram] [4C] [Range of strain rate related to the region] [Area Number] [Strain Rate Range] [(] [s,-1 , ] [)] [Example Colors] 331 18.0–19.1 red 332 15.8–17.9 Orange 333 13.2–15.7 yellow 334 5.5–13.1 green 335 2.5–5.4 light blue 336 0–2.4 Deep Blue

[0072] The strain rate grid shown in Table 1 is an example. The grouping and distribution of strain rates can vary depending on the range of strain rates observed during the CFD analysis. Evaluation criteria such as strain histograms, average strain rates, and peak strains can be determined from strain rate contour maps. Strain rate contour maps can be analyzed to identify regions in the mixing vessel subjected to high levels of strain.

[0073] As previously mentioned, the quantitative and qualitative evaluation criteria (e.g., evaluation standards) of a hybrid criterion can be determined using CFD analysis. The evaluation criteria determined by CFD analysis correspond to a set of test values ​​for the hybrid criterion parameters. The relationship between the evaluation criteria and the corresponding hybrid criterion parameters can be used to assess the impact of changes in the hybrid criterion parameters on the overall utility and / or effectiveness of the hybrid criterion.

[0074] Referring again to Figure 2, the method 200 for developing a predictive model for evaluating a hybrid criterion may include constructing candidate predictive models 204. For example, candidate predictive models may be constructed after determining evaluation criteria for the corresponding test values ​​of the hybrid criterion parameters. For each validated evaluation criterion, one or more candidate predictive models may be selected.

[0075] Figure 5 illustrates an exemplary method for developing and ranking potential predictive models for evaluation criteria. The method may include developing a domain of applicable models (step 401), removing duplicate models and models with variance inflation factors greater than or equal to a collinearity threshold (step 402), identifying a pool of candidate models (step 403), and ranking the candidate models based on complexity and relevance (step 404). This method for developing and ranking potential predictive models can be applied to each evaluation criterion identified in the DOE design to generate a ranked pool of candidate predictive models for each evaluation criterion.

[0076] Developing and ranking potential predictive models involves developing a domain of applicable models for a given evaluation criterion based on validated mixed criterion parameters. In this case, a model refers to an algebraic expression that relates the evaluation criterion to the mixed criterion parameters. When developing the domain of applicable models, univariate, bivariate, trivariate, and other multivariate relationships between the mixed criterion parameters are considered. For example, products, quotients, exponentials, and other multivariate relationships between the mixed criterion parameters may be considered. The domain of applicable models may also include known mechanical or experimental relationships. In some implementations, the domain of applicable models includes tens of thousands of models, such as over 50,000 potential predictive models.

[0077] After developing the model domain, models with duplicate parameters can be removed. For example, when developing algebraic expressions that correlate evaluation criteria with mixed criterion parameters, equivalent expressions can be created. These equivalent expressions can functionally represent duplicates that can be removed from the domain. The variance inflation factor (VIF) of each remaining model can be calculated, and models with a VIF greater than or equal to a collinearity threshold can be removed from the domain. In some implementations, the collinearity threshold is four or fewer, such as two, three, or four. After removing models with duplicate parameters and models with a VIF greater than or equal to the collinearity threshold, the remaining subset of models can include hundreds of models. For example, the remaining subset of models can include fewer than or equal to 500 models.

[0078] A pool of candidate predictive models can be identified from the remaining subset of models. For example, the pool may include univariate, bivariate, and trivariate models. Each candidate predictive model from the pool may have an R² value greater than or equal to approximately 0.70. In some implementations, each candidate predictive model from the pool may have an R² value greater than or equal to approximately the following: 0.60, 0.70, 0.75, 0.80, 0.85, 0.9, or 0.95. In some implementations, the pool includes univariate models with the largest R² value, bivariate models with the largest R² value, and trivariate models with the largest R² value. After identifying the pool of candidate predictive models, they can be ranked according to complexity and relevance. For example, predictive models with higher relevance to data obtained from CFD analysis (e.g., higher R² values) may be ranked higher based on relevance, while predictive models with lower complexity (e.g., fewer terms) may be ranked higher based on complexity. The following examples describe further examples of evaluating the relevance of potential predictive models to CFD analysis results. These two ranking methods can be combined, resulting in a predictive model with higher relevance to the data obtained from the CFD analysis (e.g., a higher R² value) and lowest complexity (e.g., fewer terms) being ranked higher than a predictive model with a lower R² value and / or higher complexity.

[0079] After ranking the prediction models according to complexity and relevance, a prediction model can be selected based on the desired complexity and relevance attributes. The selected prediction model can be further studied using test values ​​of other mixed criterion parameters or mixed criterion types.

[0080] Referring again to Figure 2, the method 200 for developing a predictive model for evaluating a hybrid criterion may include evaluating the prediction. Candidate predictive models may be tested within a range of increasing test values ​​of the hybrid criterion parameters to generate evaluation criteria for the predictions.

[0081] The predicted evaluation criteria can be compared with a comprehensive survey or CFD analysis to validate the selected predictive model. Once the predictive model has been validated against the evaluation criteria and a set of mixture criteria parameters, it can be used to evaluate thousands of mixture criteria in a high-throughput manner. The rate at which mixture criteria can be evaluated using the predictive model resolves the optimal mixture criterion parameter conditions.

[0082] Alternatively, known mechanical or experimental relationships from the literature can be compared with candidate prediction models. If the relevance of a candidate prediction model is improved by adding a term from a known or experimental relationship, that term can be incorporated into the candidate prediction model.

[0083] As more predictive models are generated and the CFD analysis database grows, more accurate predictive models will be produced for each validated evaluation criterion. All evaluation criteria for all mixed criterion parameters can be evaluated in the high-throughput manner described above to determine which mixed criterion parameters lead to sufficient evaluation criteria.

[0084] Advantageously, the high-throughput approach to evaluating blending criteria can significantly reduce the time required to validate blending criteria applicable to the production of biopharmaceutical products. [, Example , ]

[0085] [Example] [1]

[0086] CFD mixing time analysis was performed using mixing time as the evaluation criterion, and batch size, impeller speed, and solution viscosity were determined as mixing criterion parameters. A candidate prediction model was identified and described by Equation 1: T blend = c1 + c2X1 + c3X3 + c4X2X3 Equation (1) Where Tblend is the mixing time, X1 is the batch size, X2 is the impeller speed, X3 is the solution viscosity, and c1, c2, c3, and c4 are constants. For the tested values ​​of the mixing criterion parameters, a graph was plotted between the Tblend determined by CFD and the Tblend determined by Equation 1, as shown in Figure 6. Figure 6 also shows the 1:1 correlation line and the area surrounding the 1:1 correlation line to illustrate the correlation between the predictive model and the CFD analysis results. Compared to known relationships from Flickinger and Nienow, Scale-Up, Stirred Tank Reactors, Encyclopedia of Industrial Biotechnology (2010), the predictive model shows a better correlation with the CFD analysis.

[0087] [Example] [2]

[0088] Using average strain rate as the evaluation criterion, batch size, impeller speed, and solution viscosity were determined as mixing criterion parameters, and CFD strain contour analysis was performed. A candidate prediction model was identified and described by Equation 2: Equation (2)

[0089] Where γmean is the average strain rate, X1 is the batch size, X2 is the impeller velocity, and c1, c2, and c3 are constants. For the test values ​​of the hybrid criterion parameters, a graph is plotted between the γmean of the test values ​​of the hybrid criterion parameters determined by CFD and the γmean determined by Equation 2, and it is shown in Figure 7. Figure 7 also shows the 1:1 correlation line and the area around the 1:1 correlation line to illustrate the correlation between the prediction model and the CFD analysis results. Based on Ladner et al., CFD Supported Investigation of Shear Induced by Bottom-Mounted Magnetic Stirrer in Monoclonal Antibody Formulation, Pharm. Res. 35(11): 215, September 25, 2018., by adding This improves the relevance of the prediction model.

[0090] [Example] [3]

[0091] Strain rate histograms can be generated using CFD analysis. However, generating a strain rate histogram for a combination of test values ​​is time-intensive. A more efficient approach could involve generating a predictive model to describe the cumulative strain and plotting the strain rate histogram based on the predictive model. An example of a strain rate histogram generated using a predictive model is shown in Figure 8.

[0092] Referring to Figure 8, points for the strain rate histogram (e.g., points at t=20, t=40, t=60, t=75, t=80, and t=90) are generated based on the strain rate prediction model (Equation 2). These points are plotted in the histogram, as shown in Figure 8. Compared to traditional CFD-based exposure analysis, the cumulative strain described by the histogram can be generated faster and with less associated labor.

[0093] [Example] [4]

[0094] Without being constrained by theory, possible mechanisms for the formation of visible and subvisible particles are shown in Figures 9A to 9C. Individual proteins 702 (e.g., host cell proteins, proteins of interest, etc.) may be present in solution 700 within mixing container 100. As shown in Figure 9A, the surface 710 of solution 700 may initially be free of protein aggregates 712.

[0095] Protein 702 can deform in response to surface tension adsorbed at a gas-liquid interface (e.g., surface 710). During deformation, charged regions of protein 702 can be exposed. Due to the thermodynamic environment, the exposed charged regions may aggregate. The aggregated protein can form a network 712 at surface 710, as shown in Figure 9B. When the surface 710 of solution 700 is disturbed (e.g., due to mixing criteria), the network 712 may be disrupted, and fragments of the disrupted network 712 may be drawn into the bulk of solution 700.

[0096] The fragmented network can aggregate with other proteins 702 to form larger networks 712, which will then break down again and be absorbed into the bulk of the solution 700. When the fragments of the protein network reach a sufficiently large size, they are detected as large aggregates 720 within the solution 700, as shown in Figure 9C. Large aggregates 720 may appear as visible particles and can cause turbidity in the solution 700.

[0097] Traditional approaches to addressing particle formation risk rely on studies of hydrodynamic shear. However, hydrodynamic shear does not account for protein aggregate formation, and shear-based proportional testing cannot predict evaluation criteria for all production scales. The risk of particle formation remains an obstacle to the development of mixing criteria due to the difficulty in quantifying the impact of particle formation, variations in filter performance, and a lack of understanding of the long-term behavior of visible and subvisible particles in solution.

[0098] Stress at the gas-liquid interface is likely the primary factor in aggregate formation. Air inclusions also contribute to aggregate formation. Surface tension and free energy estimates, as well as atomic force microscopy observations, support the role of air inclusions in aggregate formation. Solid-liquid interface stress, cavitation erosion, nucleation, and thermal stress may also play a secondary role in aggregate formation.

[0099] To better quantify the risk of particle formation, a predictive model can be developed according to the embodiments described herein, which describes the risk of particle formation as a function of a mixture of criterion parameters. Possible mixture of criterion parameters include the characteristics of the aggregate protein, the excipient properties of the solution, and environmental factors (e.g., temperature, pressure, etc.).

[0100] CFD analysis can determine vertical velocity contours and volumetric average velocities. Figure 10A shows an example of vertical velocity contours determined by CFD. Figure 10B shows an example of volumetric average velocities determined by CFD. In this paper, volumetric average velocity refers to the spatially average fluid velocity within a volume near the liquid surface.

[0101] Similar to the strain rate contour plot described above (Figure 4C), in practice, different regions of the vertical velocity contour plot and the volume average velocity can be represented by different colors assigned according to the velocity of the region. Table 2 shows exemplary colors that can be used to represent the regions of varying vertical velocity shown in Figure 10A and the regions of varying volume average vertical velocity shown in Figure 10B. [surface] [2] [-with diagram] [10A] [and] [10B] [Example colors related to the region] [Area Number] [Example Colors] 331 red 332 Orange 333 yellow 334 green 335 light blue 336 Deep Blue

[0102] The relationship between vertical velocity contour lines determined by CFD and their positions within the mixing container can be plotted to illustrate the relationship between relative differences in vertical velocity as a function of position. For example, CFD analysis can determine the vertical velocity value for each computational cell in the mixing container. The relationship between vertical velocity and linear displacement along a radius from the center of the mixing container can be plotted, as shown in Figure 11. Each measurement 801 corresponds to a computational cell having a vertical velocity and a position along the radius of the mixing container. A weighted average 820 can be determined from each measurement 801, where the weight assigned to each measurement 801 is related to the volume of the computational cell corresponding to that measurement 801.

[0103] The aforementioned techniques can be used to develop predictive models for assessing the risk of aggregate formation. For example, the characteristics of aggregate proteins, the excipient properties of the solution, the relationship between environmental factors and aggregate formation can be determined using vertical velocity contours or volume-averaged velocities derived from CFD.

[0104] This disclosure is further described by the following non-limiting items.

[0105] Project 1: A method for developing predictive models, the method comprising: (a) Identify the mixed criterion parameters used in the prediction model; (b) Select the test value for this hybrid criterion parameter; (c) Perform computational fluid dynamics (CFD) simulations for each combination of test values; (d) Generate a potential prediction model domain related to the parameters of the hybrid criterion; and (e) Sort the potential prediction model domains associated with the mixed criterion parameters.

[0106] Project 2: The method of Project 1 further includes: Following step (a), the evaluation criteria used for the predictive model are identified; Identify the CFD simulations required to generate the evaluation criteria after step (b); Following step (d), a pool of candidate predictive models is identified from the potential predictive model domain; and Sort the pool of candidate prediction models.

[0107] Project 3: The method of any one of Project 1 or 2, wherein the mixing criterion parameters include two or more of the following: impeller speed, batch size, solution viscosity, solution density, mixing container size, and mixing container geometry.

[0108] Project 4: The method of Project 1, wherein the evaluation criteria include two or more of the following: flow pattern, fluid velocity distribution, fluid flow vector field, fluid flow streamlines, steady-state mixing time, transient mixing time, residence time distribution, contour shear strain rate, mean shear strain rate, exposure analysis, and power consumption.

[0109] Project 5: The method of any one of Project 2 or 4, wherein the validated CFD simulation includes steady-state flow analysis, transient flow analysis, mixing time analysis and / or exposure analysis.

[0110] Project 6: The method of Project 2 further includes, after generating the potential predictive model domain and before confirming the pool of candidate predictive models, Calculate the variance inflation factor for each potential prediction model in the potential prediction model domain; and Remove potential predictive models from the potential predictive model domain that have a variance inflation factor greater than or equal to the collinearity threshold, thereby generating a subset of potential predictive models.

[0111] Project 7: The method of Project 6, wherein the candidate predictive model pool includes univariate models from the subset whose R² values ​​are higher than all other univariate models in the subset, and bivariate models from the subset whose R² values ​​are higher than all other bivariate models in the subset.

[0112] Project 8: The method of Project 2, wherein sorting the candidate prediction model pool includes sorting the candidate prediction model pool based on the number of items, sorting the candidate prediction model pool based on the R2 value, or both.

[0113] Item 9: A method as described in any of Items 1 to 8, wherein the test value is the first test value, and the method further includes: Using candidate prediction models from the pool of candidate prediction models, an estimate of the evaluation criterion corresponding to the combination of the second test values ​​is generated.

[0114] Item 10: The method of Item 9, wherein the method further includes: The CFD simulation is performed on the combination of the second test values ​​to generate the evaluation criteria corresponding to the combination of the second test values; and The evaluation criterion corresponding to the combination of the second test values ​​is compared with the estimated value of the evaluation criterion corresponding to the combination of the second test values.

[0115] Project 11: A method for developing predictive models, the method comprising: Identify the parameters of the first, second, and third hybrid criteria used in this prediction model; Identify the first and second evaluation criteria used for this predictive model; Select the first test value for the first hybrid criterion parameter; Select the second test value for the second mixing criterion parameter; Select the third test value for this third hybrid criterion parameter; Identify the first computational fluid dynamics (CFD) simulation required to generate this first evaluation criterion; Identify the second CFD simulation required to generate the second evaluation criterion; By performing the first CFD simulation on each combination of the first test value, the second test value, and the third test value, a first evaluation criterion corresponding to each combination of the first test value, the second test value, and the third test value is generated; By performing the second CFD simulation on each combination of the first test value, the second test value, and the third test value, a second evaluation criterion corresponding to each combination of the first test value, the second test value, and the third test value is generated; Generate a first domain of a first prediction model that associates the first, second, and third mixed criterion parameters with the first evaluation criterion; and A second domain is generated that associates the parameters of the first, second, and third mixed criteria with the second evaluation criterion.

[0116] Item 12: The method of Item 11 further includes: Calculate the variance inflation factor for each first prediction model and each second prediction model; The first subset of the first prediction models is generated by removing the first domain of the first prediction model from the first prediction model with a variance inflation factor greater than or equal to three. The second subset of the second prediction models is generated by removing the second domain of the second prediction model from the second prediction model with a variance inflation factor greater than or equal to three. A first pool of candidate first predictive models is identified, comprising univariate models whose R² values ​​are higher than all other univariate models in the first subset, bivariate models whose R² values ​​are higher than all other bivariate models in the first subset, and trivariate models whose R² values ​​are higher than all other trivariate models in the first subset; and A second pool of candidate second prediction models is identified, which includes univariate models from the second subset whose R² values ​​are higher than all other univariate models in the second subset, bivariate models from the second subset whose R² values ​​are higher than all other bivariate models in the second subset, and trivariate models from the second subset whose R² values ​​are higher than all other trivariate models in the second subset.

[0117] Item 13: The method of Item 12 further includes: Select the fourth test value for the parameter of this first mixing criterion; Select the fifth test value for this second mixing criterion parameter; Select the sixth test value for this third hybrid criterion parameter; Using each candidate first prediction model in the first pool of candidate first prediction models, an estimated first evaluation criterion is generated corresponding to each combination of the fourth, fifth, and sixth test values; By performing the first CFD simulation on each combination of the fourth, fifth, and sixth test values, a first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values ​​is generated; and The estimated first evaluation criterion generated by each candidate first prediction model in the first pool of candidate first prediction models is compared with the first evaluation criterion corresponding to each combination of the fourth, fifth and sixth test values.

[0118] Item 14: The method of Item 13 further includes: Using each candidate second prediction model in the second pool of candidate second prediction models, an estimated second evaluation criterion is generated corresponding to each combination of the fourth, fifth, and sixth test values; By performing the second CFD simulation on each combination of the fourth, fifth, and sixth test values, a second evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values ​​is generated; and The estimated second evaluation criterion generated by each candidate second prediction model in the second pool of candidate second prediction models is compared with the second evaluation criterion corresponding to each combination of the fourth, fifth and sixth test values.

[0119] Item 15: The method of Item 14 further includes: Based on a comparison of the estimated first evaluation criterion with the first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values, a first prediction model is selected from the first pool of candidate first prediction models; A second prediction model is selected from the second pool of candidate second prediction models based on a comparison of the estimated first evaluation criterion with the first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values. Using the first prediction model, a first evaluation criterion corresponding to the hybrid criterion is determined; and Using this second prediction model, a second evaluation criterion corresponding to the hybrid criterion is determined.

[0120] Item 16: The method of Item 9, wherein the first and second evaluation criteria are selected from the list including: flow pattern, fluid velocity distribution, fluid flow vector field, fluid flow streamline, steady-state mixing time, transient mixing time, residence time distribution, contour shear strain rate, mean shear strain rate, exposure analysis, and power consumption.

[0121] Project 17: A method for modeling shear strain associated with hybrid criteria, the method comprising: Identify the mixed criterion parameters used in the prediction model; Select the test value for this hybrid criterion parameter; Computational fluid dynamics exposure analysis is performed on each combination of test values ​​to generate the shear strain corresponding to each combination of test values; Confirm the candidate prediction model pool; Sort the pool of candidate prediction models; Select a prediction model from the pool of candidate prediction models; and Using this prediction model, the cumulative shear strain of the hybrid criterion is evaluated at multiple time intervals to generate shear strain histogram data.

[0122] Item 18: The method of Item 17, wherein the mixing criterion parameters include two or more of the following: impeller speed, batch size, solution viscosity, solution density, mixing container size, and mixing container geometry.

[0123] Project 19: The method of Project 17, wherein the mixing criterion is a mixing criterion associated with the biopharmaceutical product in the bioreactor.

[0124] Project 20: The method of Project 17 further includes using the shear strain histogram data to assess the risk of visible or subvisible particle formation.

[0125] Project 21: Similar to the method in Project 17, where sorting the candidate prediction model pool includes sorting the candidate prediction model pool based on the number of terms, sorting the candidate prediction model pool based on the R² value, or both; and Selecting a predictive model from the pool of candidate predictive models includes selecting the model with the highest R² value.

[0126] Those skilled in the art will understand that the concepts upon which this disclosure is based can be readily used as the basis for designing other methods and systems for achieving the various objectives of this disclosure. Therefore, the scope of the claims should not be considered limited by the foregoing description.

[0127] 100: Mixing container 110: Impeller 200: Developing methods for evaluating predictive models based on hybrid criteria 201: Mapping Design Space 202: Constructing Experimental Design 203: Perform computational fluid dynamics (CFD) analysis 204: Constructing Candidate Prediction Models 205: Assessment and Forecast 310: Vector 320: Streamline 331: Marked area 332: Marked area 333: Marked area 334: Marked area 335: Marked area 336: Marked area 401: Steps 402: Steps 403: Steps 404: Steps 605: Area 610: Curve 700: solution 702: Protein 710: Surface 712: Mesh 720: Large Aggregate 801: Measured value 820: Weighted average

Claims

1. A method for developing a predictive model, characterized in that it includes: (a) Identify the mixing criterion parameters used for the prediction model, wherein the mixing criterion parameters include two or more of the following: impeller speed, batch size, solution viscosity, solution density, mixing container size, and mixing container geometry; (b) Select test values ​​for the mixing criterion parameters; (c) Perform computational fluid dynamics (CFD) simulations on each combination of test values; (d) Generate potential prediction model domains associated with the mixing criterion parameters; and (e) Sort the potential prediction model domains associated with the mixing criterion parameters.

2. The method as described in claim 1, further comprising: After step (a), identify the evaluation criteria for the prediction model; identify the CFD simulations to be performed in order to generate the evaluation criteria after step (b); identify a pool of candidate prediction models from the potential prediction model domain after step (d); and sort the pool of candidate prediction models.

3. The method as described in claim 2, wherein the evaluation criteria include two or more of the following: flow pattern, fluid velocity distribution, fluid flow vector field, fluid flow streamlines, steady-state mixing time, transient mixing time, residence time distribution, contour shear strain rate, mean shear strain rate, exposure analysis, and power consumption.

4. The method as described in request item 2 or 3, wherein the confirmed CFD simulation includes steady-state flow analysis, transient flow analysis, mixing time analysis and / or exposure analysis.

5. The method as described in claim 2, further comprising, after generating the potential predictive model domain and before confirming the pool of candidate predictive models, calculating the variance inflation factor of each potential predictive model in the potential predictive model domain; and removing potential predictive models from the potential predictive model domain whose variance inflation factor is greater than or equal to a collinearity threshold, thereby generating a subset of potential predictive models.

6. The method as described in claim 5, wherein the candidate predictive model pool includes univariate models from the subset whose R2 values ​​are higher than all other univariate models in the subset, and bivariate models from the subset whose R2 values ​​are higher than all other bivariate models in the subset.

7. The method described in Request 2, wherein sorting the candidate prediction model pool includes sorting the candidate prediction model pool based on the number of items, sorting the candidate prediction model pool based on the R2 value, or both.

8. The method as described in any of claims 2, 3, 6, or 7, wherein the test value is a first test value, and the method further includes: Using candidate prediction models from the pool of candidate prediction models, an estimate of the evaluation criterion corresponding to the combination of the second test values ​​is generated.

9. The method as described in claim 8, wherein the method further comprises: The CFD simulation is performed on the combination of the second test values ​​to generate an evaluation criterion corresponding to the combination of the second test values; And compare the evaluation criterion corresponding to the combination of the second test values ​​with the estimated value of the evaluation criterion corresponding to the combination of the second test values.

10. The method as described in claim 2, wherein the prediction model is an algebraic expression that associates the evaluation criterion with the mixed criterion parameters.

11. The method as described in claim 2, wherein the results of the CFD simulation include vector plots, streamline plots, strain rate contour plots, strain histograms, flow patterns, fluid velocity distribution, fluid flow vector field, fluid flow streamlines, steady-state mixing time, transient mixing time, residence time distribution, contour shear strain rate, mean shear strain rate, exposure analysis, power consumption, pressure, turbulence dissipation rate, and / or Komogorov length.

12. The method as described in claim 2, wherein the pool of candidate prediction models is configured to quantify the risk of gas-liquid interfacial stress, air inclusions, or visible or subvisible particle formation associated with the confirmed mixing criterion parameters.

13. The method as described in claim 1, wherein the mixed criterion parameter includes a first mixed criterion parameter, a second mixed criterion parameter, and a third mixed criterion parameter, wherein the test value includes a first test value for the first mixed criterion parameter, a second test value for the second mixed criterion parameter, and a third test value for the third mixed criterion parameter, and the method further includes: Following step (a), the first evaluation criterion and the second evaluation criterion used for the prediction model are confirmed; Following step (b), a first computational fluid dynamics (CFD) simulation is validated to generate the first evaluation criterion; following step (b), a second CFD simulation is validated to generate the second evaluation criterion; in step (c), the first CFD simulation and the second CFD simulation are performed, generating a first evaluation criterion corresponding to each combination of the first test value, the second test value, and the third test value by performing the first CFD simulation on each combination of the first test value, the second test value, and the third test value; generating a second evaluation criterion corresponding to each combination of the first test value, the second test value, and the third test value by performing the second CFD simulation on each combination of the first test value, the second test value, and the third test value; in step (d), a first domain of a first prediction model is generated that associates the first mixture criterion parameter, the second mixture criterion parameter, and the third mixture criterion parameter with the first evaluation criterion; and a second domain of a second prediction model is generated that associates the first mixture criterion parameter, the second mixture criterion parameter, and the third mixture criterion parameter with the second evaluation criterion.

14. The method as described in claim 13, further comprising: Calculate the variance inflation factor for each first prediction model and each second prediction model; The first subset of the first prediction models is generated by removing the first prediction models from the first domain of the second prediction model with a variance inflation factor greater than or equal to three; the second subset of the first prediction models is generated by removing the second prediction models from the second domain of the second prediction model with a variance inflation factor greater than or equal to three; a first pool of candidate first prediction models is identified, the first pool including univariate models from the first subset whose R2 value is higher than all other univariate models in the first subset, bivariate models from the first subset whose R2 value is higher than all other bivariate models in the first subset, and trivariate models from the first subset whose R2 value is higher than all other trivariate models in the first subset; and a second pool of candidate second prediction models is identified, the second pool including univariate models from the second subset whose R2 value is higher than all other univariate models in the second subset, bivariate models from the second subset whose R2 value is higher than all other bivariate models in the second subset, and trivariate models from the second subset whose R2 value is higher than all other trivariate models in the second subset.

15. The method as described in claim 14, further comprising: Select the fourth test value for the parameter of the first mixing criterion; Select the fifth test value for this second hybrid criterion parameter; Select a sixth test value for the third mixture criterion parameter; use each candidate first prediction model in the first pool of candidate first prediction models to generate an estimated first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values; generate a first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values ​​by performing the first CFD simulation on each combination of the fourth, fifth, and sixth test values. And compare the estimated first evaluation criterion generated by each candidate first prediction model in the first pool of candidate first prediction models with the first evaluation criterion corresponding to each combination of the fourth test value, the fifth test value and the sixth test value.

16. The method as described in claim 15, further comprising: Using each candidate second prediction model in the second pool of candidate second prediction models, an estimated second evaluation criterion is generated corresponding to each combination of the fourth test value, the fifth test value, and the sixth test value; by performing the second CFD simulation on each combination of the fourth test value, the fifth test value, and the sixth test value, a second evaluation criterion corresponding to each combination of the fourth test value, the fifth test value, and the sixth test value is generated; and the estimated second evaluation criterion generated by each candidate second prediction model in the second pool of candidate second prediction models is compared with the second evaluation criterion corresponding to each combination of the fourth test value, the fifth test value, and the sixth test value.

17. The method as described in claim 16, further comprising: Based on the comparison of the estimated first evaluation criterion with the first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values, a first prediction model is selected from the first pool of candidate first prediction models; based on the comparison of the estimated first evaluation criterion with the first evaluation criterion corresponding to each combination of the fourth, fifth, and sixth test values, a second prediction model is selected from the second pool of candidate second prediction models; using the first prediction model, a first evaluation criterion corresponding to the mixed criterion is determined; and using the second prediction model, a second evaluation criterion corresponding to the mixed criterion is determined.

18. The method as described in claim 1, wherein, Based on model complexity and model correlation with CFD values, the potential predictive model domains are ranked.

19. The method as described in claim 1, wherein the mixing criterion is for use in a bioreactor, wherein the bioreactor includes dupilumab.

20. The method as described in claim 1, wherein the mixing criterion is for use in a bioreactor, wherein the bioreactor comprises an anti-interleukin-4 receptor antibody.

21. A method for mixing biopharmaceutical products, comprising: Develop a predictive model based on request item 1; Use this predictive model to select mixed criteria; And in accordance with the mixing guidelines, the biopharmaceutical product is mixed.

Images