ABSTRACT The repertoire of large-molecule treatments continues to expand, resulting in diverse discovery and development workflows. This diversity yields a proliferation of software solutions and procedures for molecule registration, material tracking, experiment planning, data analytics, quality control, data sharing, and decision-making. Contrasting with this manual, labor intensive, and error-prone approach, we introduce the concept of a transformative solution: an integrated platform that translates this complexity into a harmonized, open architecture encompassing all workflows and hardware systems, covering the discovery process up to developability assessment. The benefits and complexities of such a platform are evident in examples spanning different use cases and maturity levels, such as developing multi-specific antibodies and antibody-drug conjugates using shared workflows or incorporating artificial intelligence for predictive and generative tasks. This review outlines state-of-the-art concepts behind a digital platform for automating and streamlining the discovery of new large-molecule treatments.

ABSTRACT Charge heterogeneity in monoclonal antibodies (mAbs), caused by post-translational modifications, remains a substantial obstacle to ensuring consistent, stable, and effective therapeutics. Conventional optimization techniques, such as one-factor-at-a-time and design of experiments, often fail to capture the complex, nonlinear interactions between culture parameters (e.g. pH, temperature, duration) and medium components (e.g. glucose, metal ions, amino acids). This review highlights machine learning (ML) as a powerful approach for modeling these relationships and forecasting charge variant profiles in CHO cell-based mAb process development. We summarize supervised learning and regression methods used to link process conditions with charge heterogeneity and present case studies showing ML’s role in reducing acidic and basic variants. We also discuss challenges related to data quality, model interpretability, scalability, and regulatory compliance. Finally, we propose a roadmap for adaptive, ML-driven optimization strategies for bioprocess development, aligned with Quality-by-Design principles.

ABSTRACT Immunogenicity prediction is widely used in the developability assessment of antibodies, and many marketed and clinical-stage therapeutics have a predicted T-cell epitope in the second complementary-determining region of their light chain (CDR2L). To investigate such CDR2Ls in more detail, we identified an antibody with a CDR2L for which a patient had developed treatment-emergent (TE) anti-drug antibodies (ADAs) in a clinical setting. With this, we establish the importance of predicted T-cell epitopes in CDR2L. In the course of deleting the T-cell epitope, we decided to aim for a solution that can be applied broadly to facilitate larger high-throughput discovery campaigns. For this purpose, we have developed a double-mutation scheme that targets AHo67 (Kabat51) and AHo68 (Kabat52) in the CDR2L. This 67G-68G mutation scheme was applied to all light chain sequences of a tri-specific single-chain diabody fused to a single-chain variable fragment (scMATCH3™) antibody for which TE ADAs had been observed. Analyses of patient sera showed that introduction of 67 G-68 G in CDR2L in combination with our previously described T101S-T146K (Kabat: T87S-T110K) framework mutations led to a scMATCH3 antibody with significantly reduced levels of both preexisting and TE ADA reactivities. For a diverse collection of single-chain variable fragments, application of the 67 G-68 G mutation scheme was experimentally seen to not substantially affect the functional or biophysical properties of the molecules, suggesting that this mutation scheme may be applicable to the improvement of therapeutic safety of antibodies of many types, with CDR2L-associated immunogenicity.

ABSTRACT Targeting various combinations of tumor antigens and immune cell receptors is of increasing importance in antibody-based cancer immunotherapy. Here, we present a novel modular P329G-engager platform that enables rapid combination of primary tumor-targeting and secondary immune effector antibodies. The platform utilizes two antibodies, each selected from: 1) a set of tumor-targeting adaptor antibodies, bearing P329G mutations in the Fc region, and 2) a set of P329G-targeting (bispecific) cell engagers, including innate and T cell engagers, costimulators and immunocytokines. Specifically, upon defining a tumor-associated cell surface target, a primary adaptor – tumor antigen-binding IgG1 antibody with Fc-silencing P329G L234A L235A mutations – is administered. Subsequently, a secondary antibody recognizing the P329G mutation is chosen from a panel of effector cell engagers with different modes of action – ADCC-competent P329G-innate cell engagers (P329G-ICE), P329G-T cell bispecifics (P329G-TCB), P329G-costimulators (P329G-CD28/4-1BBL), or P329G-immunocytokine (P329G-IL2v). In vitro assays showed that all P329G-targeting modalities induce anti-tumoral and/or immunomodulatory activity when both components were combined. In vivo, tumor shrinkage and T cell infiltration were confirmed in tumor-bearing humanized mice treated with P329G-mutated CEACAM5 adaptor IgG and P329G-TCB. Individually, neither the adaptor nor the P329G-TCB induced efficacy, validating the requirement for primary and secondary antibody assembly for T cell-engaging activity. These results provided evidence for the in vivo assembly and subsequent pharmacological activity, and provide preclinical proof-of-concept for the P329G-engager platform as an efficacious tool in drug discovery. Ultimately, this modular approach may enable mix-and-match drug assembly as a novel therapeutic principle in immunotherapy.

ABSTRACTGrowing knowledge around disease states has led to opportunities within research to make designer molecules with improved specificity and broader efficacy. These next-generation molecules frequently take advantage of multispecific targeting and controlled mechanisms of action by utilizing four unique peptide chains as seen in many bispecific or trispecific antibodies. However, with all the opportunities these multispecifics offer, their increased biological complexities come with increased challenges during expression and purification to produce high-quality material. Lower yields accompanied with a high degree of mispairing after the initial capture purification step are often limiting factors. Developing new methods for stable pool expression can offer a strong advantage for progressing these molecules through research toward development. Here, we implemented optimized stable cell pools using targeted dual selection (TDS), a novel approach that combines specified selective pressure with transposon-guided semi-targeted gene integration. By utilizing key analytical data obtained during early-stage high-throughput transient productions, we can predict improved vector configurations for the generation of optimized TDS stable pools. We demonstrate that this design can improve molecule quality at the initial capture purification step in two Y-shaped bispecific molecules and two cross-over dual variable trispecific molecules by achieving up to four-fold increase in protein of interest yields while maintaining product quality. Use of this strategy in research can enable simplified purification strategies as well as increased production yields required for successful and timely project progression.

ABSTRACT Messenger RNA (mRNA) has emerged as a powerful tool for protein expression in clinical settings, yet its potential as a platform for biologics manufacturing remains underexplored. Here, we evaluate transient mRNA transfection in Chinese hamster ovary (CHO) cells as a rapid and versatile system for protein production. Using reporter mRNAs, we optimize transfection efficiency and benchmark performance against industry-standard plasmid transfection and stable cell line methods. We demonstrate that co-transfection of heavy and light chain mRNAs enables the efficient synthesis, assembly and secretion of the monoclonal antibody bevacizumab with high fidelity. Compared to conventional approaches, mRNA transfection drives rapid and predictable protein expression, reducing cell incubation times and enabling sequential or conditional expression. These features highlight mRNA as a flexible and efficient platform for transient expression, providing a foundation for accelerating the development and manufacturing of biologics.

ABSTRACT Chimeric antigen receptor (CAR)-modified T cells have garnered substantial attention due to their clinical success, culminating in six Food and Drug Administration-approved therapies for hematological malignancies. Notably, CD19-specific CAR T cell therapies have achieved remarkable clinical efficacy in treating B-cell malignancies, but these profound and durable responses are not observed in CAR T therapies targeting other indications, particularly solid tumors. Key design elements of CAR constructs – namely, antigen binding affinity and spacer length – play critical roles in determining T cell effector function and overall therapeutic effectiveness. Refining CAR designs may enhance T cell functionality, extend clinical application, and potentially apply CAR T cell therapies across a wider array of malignancies. In this study, affinity variant and spacer variant CARs targeting BCMA and DLL3 tumor antigens were evaluated using in vitro measurements of antigen-binding properties and effector function. Each panel of CARs spanned 2–3 logs of antigen binding affinity (BCMA: 181 pM KD to 74 nM KD, DLL3: 417 pM to 407 nM). Additionally, CAR T cells were challenged with tumor spheroids composed of BCMA+ H929 and DLL3+ SHP77 tumor cells. We show that for both tumor models, higher affinity CARs (KD stronger than approximately 100 nM) paired with an intermediate length spacer (IgG1 Fc, CH2-CH3, 230AA) elicited the strongest levels of tumor killing, CAR+ T cell expansion, and proinflammatory cytokine production. These CARs displayed the strongest cellular affinity when measured in a conjugation assay, suggesting a relationship between cellular affinity and T cell functional performance. This study highlights the critical role of CAR design in enhancing T cell functionality, demonstrating that high-affinity CARs combined with intermediate-length spacers yield superior performance in targeting BCMA and DLL3 antigens. This study provides a framework for rational CAR design, informing strategies to broaden the clinical utility of CAR T-cell therapies beyond hematologic cancers.

ABSTRACT T-cell receptor mimic (TCRm) antibodies are an emerging class of tumor-targeting agents used in advanced immunotherapies such as bispecific T-cell engagers and CAR-T cells. Unlike conventional antibodies, TCRms are designed to recognize peptide – human leukocyte antigen (pHLA) complexes that present intracellular tumor-derived peptides on the cell surface. Due to the typically low surface abundance and high sequence similarity of pHLAs, TCRms require high affinity and exceptional specificity to avoid off-target toxicity. Conventional methods for off-target identification such as sequence similarity searches, motif-based screening, and structural modeling focus on the peptide and are limited in detecting cross-reactive peptides with little or no sequence homology to the target. To address this gap, we developed EpiPredict, a TCRm-specific machine learning framework trained on high-throughput kinetic off-target screening data. EpiPredict learns an antibody-specific mapping from peptide sequence to binding strength, enabling prediction of interactions with unmeasured pHLA sequences, including sequence-dissimilar peptides. We applied EpiPredict to two distinct TCRms targeting the cancer-testis antigen MAGE-A4. The model successfully predicted multiple off-targets with minimal sequence similarity to the intended epitope, many of which were experimentally validated via T2 cell binding assays. These findings establish EpiPredict as a valuable tool for lead optimization of TCRms, enabling the identification of antibody-specific off-targets beyond the scope of traditional peptide-centric methods and supporting the preclinical de-risking of TCRm-based therapies.

ABSTRACT We introduce Ibex, a pan-immunoglobulin structure prediction model for antibodies, nanobodies, and T-cell receptors. Unlike previous approaches, Ibex explicitly distinguishes between bound and unbound protein conformations by training on labeled apo and holo structural pairs, enabling accurate prediction of both states at inference time. Ibex achieves state-of-the-art accuracy, demonstrating superior out-of-distribution performance on a comprehensive benchmark of high-resolution antibody structures with a mean CDR H3 RMSD of 2.28 Å. Ibex combines this accuracy with significantly reduced computational requirements, providing a robust foundation for accelerating large molecule design and therapeutic development.

ABSTRACT Conventional antibody discovery methods, such as hybridoma and phage display, face inherent limitations. Hybridoma technology relies on labor-intensive cell fusion and clone screening, often taking several weeks to obtain stable clones. Phage display allows in vitro selection but disrupts natural heavy- and light-chain pairing, potentially affecting antibody stability, safety, and developability. Single-cell approaches provide direct access to naturally paired sequences, enabling faster identification of functional candidates. Among B cell subsets, plasma cells, as terminally differentiated antibody-secreting cells, produce higher-affinity and more mature antibodies than memory B cells, yet their efficient antigen-specific enrichment at industrially relevant throughput remains challenging. Here, we present AbDrop, a microfluidics-enabled platform integrating high-throughput plasma cell capture, repertoire-level bioinformatics, and scalable antibody expression, with optional epitope binning. This workflow can process and enrich 1–2 million plasma cells per run, enabling recovery of hundreds to thousands of unique antibody sequences within a week and rapid functional validation – including binding specificity and, when performed, epitope classification – within three to four weeks. Compared with existing plasma cell – focused platforms, such as Beacon, AbDrop achieves substantially higher throughput while maintaining transparent sequence recovery and rapid downstream expression. As a proof-of-concept, we applied AbDrop to the PD-1 immune repertoire, identifying multiple functional antibodies with diverse activities, including blockers and agonists. These results demonstrate that AbDrop provides an industrially compatible, high-throughput framework for accelerated discovery and functional characterization of therapeutic antibodies.