Protein-Centric Analysis of Personalized Antibody Repertoires Using LC-MS-Based Fab-Profiling on a timsTOF.

Endogenous antibodies, or immunoglobulins (Igs), abundantly present in body fluids, represent some of the most challenging samples to analyze, largely due to the immense variability in their sequences and concentrations. It has been estimated that our body can produce billions of different Ig proteins with different isotypes, making their individual analysis seemingly impossible. However, recent advances in protein-centric proteomics using LC-MS coupled to Orbitrap mass analyzers to profile intact Fab fragments formed by selective cleavage at the IgG-hinge revealed that IgG repertoires may be less diverse, albeit unique for each donor. Serum repertoires seem to be dominated by a few hundred clones that cumulatively make up 50-95% of the total IgG content. Enabling such analyses required careful optimization of the chromatography and mass analysis, as all Fab analytes are highly alike in mass (46-51 kDa) and sequence. To extend the opportunities of this mass-spectrometry-based profiling of antibody repertoires, we here report the optimization and evaluation of an alternative MS platform, namely, the timsTOF, for antibody repertoire profiling. The timsTOF mass analyzer has gained traction in recent years for peptide-centric proteomics and found wide applicability in plasma proteomics, affinity proteomics, and HLA peptidomics, to name a few. However, for protein-centric analysis, this platform has been less explored. Here, we demonstrate that the timsTOF platform can be adapted to perform protein-centric LC-MS-based profiling of antibody repertoires. In a side-by-side comparison of the timsTOF and the Orbitrap we demonstrate that the extracted serum antibody repertoires are alike qualitatively and quantitatively, whereby in particular the sensitivity of the timsTOF platform excels. Future incorporation of advanced top-down capabilities on the timsTOF may make this platform a very valuable alternative for protein-centric proteomics and top-down proteomics and thus also for personalized antibody repertoire profiling.

A comparative analysis of recombinant Fab and full-length antibody production in Chinese hamster ovary cells.

Monoclonal antibodies are the leading class of biopharmaceuticals in terms of numbers approved for therapeutic purposes. Antigen-binding fragments (Fab) are also used as biotherapeutics and used widely in research applications. The dominant expression systems for full-length antibodies are mammalian cell-based, whereas for Fab molecules the preference has been an expression in bacterial systems. However, advances in CHO and downstream technologies make mammalian systems an equally viable option for small- and large-scale Fab production. Using a panel of full-length IgG antibodies and their corresponding Fab pair with different antigen specificities, we investigated the impact of the IgG and Fab molecule format on production from Chinese hamster ovary (CHO) cells and assessed the cellular capability to process and produce these formats. The full-length antibody format resulted in the recovery of fewer mini-pools posttransfection when compared to the corresponding Fab fragment format that could be interpreted as indicative of a greater overall burden on cells. Antibody-producing cell pools that did recover were subsequently able to achieve higher volumetric protein yields (mg/L) and specific productivity than the corresponding Fab pools. Importantly, when the actual molecules produced per cell of a given format was considered (as opposed to mass), CHO cells produced a greater number of Fab molecules per cell than obtained with the corresponding IgG, suggesting that cells were more efficient at making the smaller Fab molecule. Analysis of cell pools showed that gene copy number was not correlated to the subsequent protein production. The amount of mRNA correlated with secreted Fab production but not IgG, whereby posttranscriptional processes act to limit antibody production. In summary, we provide the first comparative description of how full-length IgG and Fab antibody formats impact on the outcomes of a cell line construction process and identify potential limitations in their production that could be targeted for engineering increases in the efficiency in the manufacture of these recombinant antibody formats.

Multiattribute Monitoring of Antibody Charge Variants by Cation-Exchange Chromatography Coupled to Native Mass Spectrometry.

The aim of this study was to characterize the product variants of a therapeutic T-cell bispecific humanized monoclonal antibody (TCB Mab, ∼200 kDa, asymmetric) and to develop an online cation-exchange chromatography native electrospray mass spectrometry method (CEC-UV-MS) for direct TCB Mab charge variant monitoring during bioprocess and formulation development. For the identification and functional evaluation of the diverse and complex TCB Mab charge variants, offline fractionation combined with comprehensive analytical testing was applied. The offline fractionation of abundant product variant peaks enabled identification of coeluting acid charge variants such as asparagine deamidation, primary and secondary Fab glycosylation (with and without sialic acid), and the presence of O-glycosylation in the G4S-linker region. Consequently, a new nonconsensus N-glycosylation motif (N-338-FG) in the heavy chain CDR region was discovered. Functional evaluation by cell-based potency testing demonstrated a clear and negative impact of both asparagine deamidations, whereas the O-glycosylation did not affect the TCB Mab biological activity. We established an online native CEC-UV-MS method, with an ammonium acetate buffer and pH gradient, to directly monitor the TCB Mab charge variants. All abundant chemical degradations and post-translational amino acid modifications already identified by offline fraction experiments and liquid chromatography mass spectrometry peptide mapping could also be monitored by the online CEC-UV-MS method. The herein reported online native CEC-UV-MS methodology represents a complementary or even alternative approach for multiattribute monitoring of biologics, offering multiple benefits, including increased throughput and reduced sample handling and intact protein information in the near-native state.

Purification of Fab and Fc using papain immobilized cryogel bioreactor separator system.

The continuous bed bioreactor systems have been used for the production of protein therapeutics, such as IgG, using immobilized enzyme in biopharmaceutical applications. We developed macroporous poly(hydroxyethyl methacrylate-co-glycidyl methacrylate) cryogel-based bioreactor matrix using sodium dodecyl sulfate as surfactans in the presence of ethylene glycol dimethacrylate as cross linking agent by bulk polymerization. The developed polyGMA immobilized bioreactor with papain enzyme was used for specific fragmentation of immunoglobulin G. The catalysis efficiency for immobilized enzyme were investigated in comparison with free enzyme. The immobilized papain displayed broad catalytic activity over a variety of conditions, with maximal activity around pH 7.0 and 70 °C. The Michaelis-Menten kinetic constant (Km), the maximum reaction velocity (Vmax), and the catalytic efficiency (kcat) for free enzyme were 0.1097 mg/mL, 29.9 mg/mL/min, and 92.01 1/min, respectively, whereas for immobilized enzyme, Km, Vmax, and kcat values were 0.1078 mg/mL, 30.53 mg/mL/min, and 94.3 1/min, respectively. In a further step, after digestion, remarkable digestion products of bioreactor, Fab and Fc fragments, produced with immobilized papain bioreactors were analyzed in two ways by SDS-PAGE and reversed-phase HPLC; it was demonstrated that papain immobilized bioreactor successfully used for the digestion of human IgG with high activity. Therefore, the polyGMA cryogel immobilized with papain exhibited a very effective matrix for the bioreactor which can be considered as an alternative bioreactor matrix with great promise in biopharmaceutical applications.

QSAR Implementation for HIC Retention Time Prediction of mAbs Using Fab Structure: A Comparison between Structural Representations.

Monoclonal antibodies (mAbs) constitute a rapidly growing biopharmaceutical sector. However, their growth is impeded by high failure rates originating from failed clinical trials and developability issues in process development. There is, therefore, a growing need for better in silico tools to aid in risk assessment of mAb candidates to promote early-stage screening of potentially problematic mAb candidates. In this study, a quantitative structure-activity relationship (QSAR) modelling workflow was designed for the prediction of hydrophobic interaction chromatography (HIC) retention times of mAbs. Three novel descriptor sets derived from primary sequence, homology modelling, and atomistic molecular dynamics (MD) simulations were developed and assessed to determine the necessary level of structural resolution needed to accurately capture the relationship between mAb structures and HIC retention times. The results showed that descriptors derived from 3D structures obtained after MD simulations were the most suitable for HIC retention time prediction with a R2 = 0.63 in an external test set. It was found that when using homology modelling, the resulting 3D structures became biased towards the used structural template. Performing an MD simulation therefore proved to be a necessary post-processing step for the mAb structures in order to relax the structures and allow them to attain a more natural conformation. Based on the results, the proposed workflow in this paper could therefore potentially contribute to aid in risk assessment of mAb candidates in early development.

Biosensing Cytokine IL-6: A Comparative Analysis of Natural and Synthetic Receptors.

Cytokines are a family of proteins which play a major role in the regulation of the immune system and the development of several diseases, from rheumatoid arthritis to cancer and, more recently, COVID-19. Therefore, many efforts are currently being developed to improve therapy and diagnosis, as well as to produce inhibitory drugs and biosensors for a rapid, minimally invasive, and effective detection. In this regard, even more efficient cytokine receptors are under investigation. In this paper we analyze a set of IL-6 cytokine receptors, investigating their topological features by means of a theoretical approach. Our results suggest a topological indicator that may help in the identification of those receptors having the highest complementarity with the protein, a feature expected to ensure a stable binding. Furthermore, we propose and discuss the use of these receptors in an idealized experimental setup.

An easy and simple separation method for Fc and Fab fragments from chicken immunoglobulin Y (IgY).

Antigen-binding (Fab) and crystallizable (Fc) fragments are the active components of yolk immunoglobulin (IgY), which have been widely used in the pharmaceutical field. However, the common purification methods for the Fab and Fc fragments use combinations of multi-columns are complex and time-consuming. The objective of this study was to improve the separation efficiency of the Fab and Fc fragments from the hydrolyzed IgY and increase the purity of the isolated Fab and Fc fragments. Natural IgY was hydrolyzed using papain for 6 hr and then treated with 45% saturated ammonium sulfate to remove small molecular-weight-peptides. The fraction containing Fab and Fc fragments was loaded on a DEAE-Sepharose ion exchange column and the Fab fraction was washed out first with 10 mM Tris-HCl buffer (pH 7.6). Then, the Fc fraction bound to the DEAE Sepharose was eluted with 10 mM Tris-HCl buffer (pH 7.6) containing 0.21 M NaCl. The purity of the two fragments was 88.7% and 90.1%, respectively. The results of Western blotting and MS analyses indicated that this method purified Fab and Fc fractions with high purity. This method is easy and simple compared with other methods, and the active fragments separated can be easily used.

Direct Determination of Antibody Chain Pairing by Top-down and Middle-down Mass Spectrometry Using Electron Capture Dissociation and Ultraviolet Photodissociation.

One challenge associated with the discovery and development of monoclonal antibody (mAb) therapeutics is the determination of heavy chain and light chain pairing. Advances in MS instrumentation and MS/MS methods have greatly enhanced capabilities for the analysis of large intact proteins yielding much more detailed and accurate proteoform characterization. Consequently, direct interrogation of intact antibodies or F(ab')2 and Fab fragments has the potential to significantly streamline therapeutic mAb discovery processes. Here, we demonstrate for the first time the ability to efficiently cleave disulfide bonds linking heavy and light chains of mAbs using electron capture dissociation (ECD) and 157 nm ultraviolet photodissociation (UVPD). The combination of intact mAb, Fab, or F(ab')2 mass, intact LC and Fd masses, and CDR3 sequence coverage enabled determination of heavy chain and light chain pairing from a single experiment and experimental condition. These results demonstrate the potential of top-down and middle-down proteomics to significantly streamline therapeutic antibody discovery.

Interlaboratory Comparison of Hydrogen-Deuterium Exchange Mass Spectrometry Measurements of the Fab Fragment of NISTmAb.

Hydrogen-deuterium exchange mass spectrometry (HDX-MS) is an established, powerful tool for investigating protein-ligand interactions, protein folding, and protein dynamics. However, HDX-MS is still an emergent tool for quality control of biopharmaceuticals and for establishing dynamic similarity between a biosimilar and an innovator therapeutic. Because industry will conduct quality control and similarity measurements over a product lifetime and in multiple locations, an understanding of HDX-MS reproducibility is critical. To determine the reproducibility of continuous-labeling, bottom-up HDX-MS measurements, the present interlaboratory comparison project evaluated deuterium uptake data from the Fab fragment of NISTmAb reference material (PDB: 5K8A ) from 15 laboratories. Laboratories reported ∼89 800 centroid measurements for 430 proteolytic peptide sequences of the Fab fragment (∼78 900 centroids), giving ∼100% coverage, and ∼10 900 centroid measurements for 77 peptide sequences of the Fc fragment. Nearly half of peptide sequences are unique to the reporting laboratory, and only two sequences are reported by all laboratories. The majority of the laboratories (87%) exhibited centroid mass laboratory repeatability precisions of ⟨ sLab⟩ ≤ (0.15 ± 0.01) Da (1σx̅). All laboratories achieved ⟨sLab⟩ ≤ 0.4 Da. For immersions of protein at THDX = (3.6 to 25) °C and for D2O exchange times of tHDX = (30 s to 4 h) the reproducibility of back-exchange corrected, deuterium uptake measurements for the 15 laboratories is σreproducibility15 Laboratories( tHDX) = (9.0 ± 0.9) % (1σ). A nine laboratory cohort that immersed samples at THDX = 25 °C exhibited reproducibility of σreproducibility25C cohort( tHDX) = (6.5 ± 0.6) % for back-exchange corrected, deuterium uptake measurements.

Top-down Mass Spectrometry Analysis of Human Serum Autoantibody Antigen-Binding Fragments.

Detecting autoimmune diseases at an early stage is crucial for effective treatment and disease management to slow disease progression and prevent irreversible organ damage. In many autoimmune diseases, disease-specific autoantibodies are produced by B cells in response to soluble autoantigens due to defects in B cell tolerance mechanisms. Autoantibodies accrue early in disease development, and several are so disease-specific they serve as classification criteria. In this study, we established a high-throughput, sensitive, intact serum autoantibody analysis platform based on the optimization of a one dimensional ultra-high-pressure liquid chromatography top-down mass spectrometry platform (1D UPLC-TDMS). This approach has been successfully applied to a 12 standard monoclonal antibody antigen-binding fragment (Fab) mixture, demonstrating the feasibility to separate and sequence intact antibodies with high sequence coverage and high sensitivity. We then applied the optimized platform to characterize total serum antibody Fabs in a systemic lupus erythematosus (SLE) patient sample and compared it to healthy control samples. From this analysis, we show that the SLE sample has many dominant antibody Fab-related mass features unlike the healthy controls. To our knowledge, this is the first top-down demonstration of serum autoantibody pool analysis. Our proposed approach holds great promise for discovering novel serum autoantibody biomarkers that are of interest for diagnosis, prognosis, and tolerance induction, as well as improving our understanding of pathogenic autoimmune processes.

Facile single-molecule pull-down assay for analysis of endogenous proteins.

The single-molecule pull-down (SiMPull) assay analyzes molecular complexes in physiological conditions from cell or tissue lysates. Currently the approach requires a lengthy sample preparation process, which has largely prevented the widespread adoption of this technique in bioanalysis. Here, we present a simplified SiMPull assay based upon dichlorodimethylsilane-Tween-20 passivation and F(ab) fragment labeling. Our passivation is a much shorter process than the standard polyethylene glycol passivation used in most single-molecule studies. The use of F(ab) fragments for indirect fluorescent labeling rather than divalent F(ab')2 or whole IgG antibodies allows for the pre-incubation of the detection antibodies, reducing the sample preparation time for single-molecule immunoprecipitation samples. We examine the applicability of our approach to recombinant proteins and endogenous proteins from mammalian cell lysates.

Assessment of quality, safety, and pre-clinical toxicity of an equine polyvalent anti-snake venom (Pan Africa): Determination of immunological cross-reactivity of antivenom against venom samples of Elapidae and Viperidae snakes of Africa.

Snakebite causes a large amount of morbidities and mortalities in Africa. The safety, efficacy, and homogeneity of anti-snake venoms are crucial for snakebite treatments to be effective with minimal adverse effects. We assessed the homogeneity of preparations of three different batches of Combipack snake venom antiserums (Pan Africa) [CSVAPA] by quantitatively analysing F(ab')2, IgG, and other contaminating proteins of plasma. LC-MS/MS analysis showed that approximately 92.4% of the proteins from the CSVAPA samples was IgG/F(ab')2 and the percent composition of contaminating proteins in CSVAPA varied from 0.07 to 4.6%. Batch 1 of the CSVAPA also contained a minor amount of undigested IgG and F(ab')2 aggregates. CSVAPA contained more than 60% venom-specific antibodies, showed moderate complement activation, no IgE contamination, safe level of endotoxin, and also showed pre-clinical safety. The immuno cross-reactivity of CSVAPA against 14 Viperidae and Elapidae snake venoms of Africa was tested by ELISA and immunoblotting, and the neutralization of major enzymatic venom activities, demonstrating that high molecular weight (>50 kDa) venom proteins are better recognized/neutralized compared to relatively low molecular weight (<20 kDa) venom proteins. CSVAPA at a dose of 3-12 times higher than the clinical dose did not cause deaths or adverse reaction of treated rabbits. The results suggest the satisfactory quality, safety, and efficacy of CSVAPA.

Quality and cost assessment of a recombinant antibody fragment produced from mammalian, yeast and prokaryotic host cells: A case study prior to pharmaceutical development.

Monoclonal antibody fragments (Fab) are a promising class of therapeutic agents. Fabs are aglycosylated proteins and so many expression platforms have been developed including prokaryotic, yeast and mammalian cells. However, these platforms are not equivalent in terms of cell line development and culture time, product quality and possibly cost of production that greatly influence the success of a drug candidate's pharmaceutical development. This study is an assessment of the humanized Fab fragment ACT017 produced from two microorganisms (Escherichia coli and Pichia pastoris) and one mammalian cell host (CHO). Following low scale production and Protein L-affinity purification under generic conditions, physico-chemical and functional quality assessments were carried out prior to economic analysis of industrial scale production using a specialized software (Biosolve, Biopharm Services, UK). Results show higher titer production when using E. coli but associated with high heterogeneity of the protein content recovered in the supernatant. We also observed glycoforms of the Fab produced from P. pastoris, while Fab secreted from CHO was the most homogeneous despite a much longer culture time and slightly higher estimated cost of goods. This study may help inform future pharmaceutical development of this class of therapeutic proteins.

Generation and selection of naïve Fab library for parasitic antigen: Anti-BmSXP antibodies for lymphatic filariasis.

Phage display has been applied successfully as a tool for the generation of monoclonal antibodies (mAbs). Naive antibody libraries are unique as they are able to overcome several limitations associated with conventional mAb generation methods like the hybridoma technology. Here, we performed an in vitro selection and generation of Fab antibodies against Brugia malayi SXP protein (BmSXP), a recombinant antigen for the detection of lymphatic filariasis. We developed a naïve multi ethnic Fab antibody library with an estimated diversity of 2.99 × 109 . The antibody library was used to screen for mAbs against BmSXP recombinant antigen. Soluble monoclonal Fab antibodies against BmSXP were successfully isolated from the naïve library. The Fab antibodies obtained were expressed and analyzed to show its binding capability. The diversity obtained from a pool of donors from various ethnic groups allowed for a diverse antibody library to be generated. The mAbs obtained were also functional in soluble form, which makes it useful for further downstream applications. We believe that the Fab mAbs are valuable for further studies and could also contribute to improvements in the diagnosis of filariasis.

Human recombinant Fab fragment from combinatorial libraries of a B-cell lymphoma patient recognizes core protein of chondroitin sulphate proteoglycan 4.

CD antigens are well known as therapeutic targets of B-cell lymphoma. To isolate therapeutic antibodies that recognize novel targets other than CD antigens, we constructed a phage display combinatorial antibody Fab library from bone marrow lymphocytes of B-cell lymphoma patient. To eliminate antibodies reactive with known B-cell lymphoma antigen, non-hematopoietic and patient's sera reactive HeLaS3 cells was selected as a target of whole cell panning. Five rounds of panning against live HeLaS3 cells retrieved single Fab clone, termed AHSA (Antibody to HeLa Surface Antigen). Using phage display random peptide library, LSYLEP was identified as an epitope sequence of AHSA. LC-MS/MS analysis of AHSA-precipitated HeLaS3 cell lysates detected several fragments corresponding to the sequence of chondroitin sulphate proteoglycan 4 (CSPG4) core protein. Since LSYLEP sequence was at the position of 313-318 of CSPG4, we considered that CSPG4 was AHSA-associated antigen. Double staining of CSPG4-postive MDA-MB-435S cells with AHSA and anti-CSPG4 rabbit antibody showed identical staining position, and reduced AHSA reactivity was observed in CSPG4-siRNA treated MDA-MB-435S cells. In conclusion, we retrieved a human Fab from antibody library of B-cell lymphoma patient, and identified CSPG4 as a recognizing antigen. AHSA may have potential benefits for development of CSPG4-targeting theranostics for B-cell lymphoma.

Quality of horse F(ab)2 antitoxins and antirabies immunoglobulins: protein content and anticomplementary activity

Background Among other applications, immunotherapy is used for the post-exposure treatment and/or prophylaxis of important infectious diseases, such as botulism, diphtheria, tetanus and rabies. The effectiveness of serum therapy is widely proven, but improvements on the immunoglobulin purification process and on the quality control are necessary to reduce the amount of protein aggregates. These may trigger adverse reactions in patients by activating the complement system and inducing the generation of anaphylatoxins. Herein, we used immunochemical methods to predict the quality of horse F(ab)2 anti-botulinum AB, anti-diphtheric, antitetanic and anti-rabies immunoglobulins, in terms of amount of proteins and protein aggregates. Methods Samples were submitted to protein quantification, SDS-PAGE, Western blot analysis and molecular exclusion chromatography. The anticomplementary activity was determined in vitro by detecting the production of C5a/C5a desArg, the most potent anaphylatoxin. Data were analyzed by one-way ANOVA followed by Tukey's post-test, and differences were considered statistically significant when p 0.05. Results Horse F(ab)2 antitoxins and anti-rabies immunoglobulin preparations presented different amounts of protein. SDS-PAGE and Western blot analyses revealed the presence of protein aggregates, non-immunoglobulin contaminants and, unexpectedly, IgG whole molecules in the samples, indicating the non-complete digestion of immunoglobulins. The chromatographic profiles of antitoxins and anti-rabies immunoglobulins allowed to estimate the percentage of contaminants and aggregates in the samples. Although protein aggregates were present, the samples were not able to induce the generation of C5a/C5a desArg in vitro, indicating that they probably contain acceptable levels of aggregates...

Sequence composition predicts immunoglobulin superfamily members that could share the intrinsically disordered properties of antibody CH1 domains.

Antibodies are central to the growing sector of protein therapeutics, and increasingly they are being manipulated as fragments and combinations. An improved understanding of the properties of antibody domains in isolation would aid in their engineering. We have conducted an analysis of sequence and domain interactions for IgG antibodies and Fab fragments in the structural database. Of sequence-related properties studied, relative lysine to arginine content was found to be higher in CH1 and CL than in variable domains. As earlier work shows that lysine is favoured over arginine in more soluble proteins, this suggests that individual domains may not be optimised for greater solubility, giving scope for fragment engineering. Across other sequence-based features, CH1 is anomalous. A sequence-based scheme predicts CH1 to be folded, although it is known that CH1 folding is linked to IgG assembly and secretion. Calculations indicate that charge interactions in CH1 domains contribute less to folded state stability than in other Fab domains. Expanding to the immunoglobulin superfamily reveals that a subset of non-antibody domains shares sequence composition properties with CH1, leading us to suggest that some of these may also couple folding, assembly and secretion.

LC-MS/MS quantification of free and Fab-bound colchicine in plasma, urine and organs following colchicine administration and colchicine-specific Fab fragments treatment in Göttingen minipigs.

Clinical evaluation of a colchicine specific antigen-binding fragment (Fab) in order to treat colchicine poisoning required the development of an accurate method allowing quantification of free and Fab-bound colchicine in plasma and urine, and free colchicine in tissues, to measure colchicine redistribution after Fab administration. Three methods have been developed for this purpose, and validated in plasma, urine and liver: total colchicine was determined after denaturation of Fab by dilution in water and heating; free colchicine was separated from Fab-bound colchicine by filtration with 30KDa micro-filters; tissues were homogenized in a tissue mixer. Deuterated colchicine was used as internal standard. Samples were extracted by liquid-liquid extraction and analyzed with a LC-MS/MS. LOQ were 0.5ng/mL in plasma and urine for free and total colchicine and 5pg/mg in tissues. The methods were linear in the 0.5-100ng/mL range in plasma and urine, and 5-300pg/mg in tissues with determination coefficients>0.99. Precision and accuracy of QC samples presented a CV<9.4%. The methods require only 200µL of sample and allow a high throughput due to short analytical run (2min). These methods were successfully applied to a pig intoxicated with colchicine and treated with colchicine specific Fab fragments.

QSAR models for prediction of chromatographic behavior of homologous Fab variants.

While quantitative structure activity relationship (QSAR) models have been employed successfully for the prediction of small model protein chromatographic behavior, there have been few reports to date on the use of this methodology for larger, more complex proteins. Recently our group generated focused libraries of antibody Fab fragment variants with different combinations of surface hydrophobicities and electrostatic potentials, and demonstrated that the unique selectivities of multimodal resins can be exploited to separate these Fab variants. In this work, results from linear salt gradient experiments with these Fabs were employed to develop QSAR models for six chromatographic systems, including multimodal (Capto MMC, Nuvia cPrime, and two novel ligand prototypes), hydrophobic interaction chromatography (HIC; Capto Phenyl), and cation exchange (CEX; CM Sepharose FF) resins. The models utilized newly developed "local descriptors" to quantify changes around point mutations in the Fab libraries as well as novel cluster descriptors recently introduced by our group. Subsequent rounds of feature selection and linearized machine learning algorithms were used to generate robust, well-validated models with high training set correlations (R2 > 0.70) that were well suited for predicting elution salt concentrations in the various systems. The developed models then were used to predict the retention of a deamidated Fab and isotype variants, with varying success. The results represent the first successful utilization of QSAR for the prediction of chromatographic behavior of complex proteins such as Fab fragments in multimodal chromatographic systems. The framework presented here can be employed to facilitate process development for the purification of biological products from product-related impurities by in silico screening of resin alternatives. Biotechnol. Bioeng. 2017;114: 1231-1240. © 2016 Wiley Periodicals, Inc.