The expanding diversity of biologic therapeutics has made the early selection of modality and molecular format a critical decision point in next generation biologics discovery. Choosing the optimal format requires a balanced consideration of anticipated functional requirements, manufacturability, and developability. Once a format is selected, thoughtful molecular design is essential to ensure successful antibody discovery, enabling both robust biological activity and favorable biophysical properties. We demonstrate how an integrated biophysical toolbox can be used to characterize antibody-antigen complex formation, and interactions, providing insights that guide format selection and molecular design while helping identify potential immunogenicity and developability risks.

Multi-specific antibodies incorporating multiple domains and linkers are susceptible to fragmentation. We investigated low molecular weight (LMW) species formation during formulation stability studies of a CHO-derived multi-specific antibody (~250 kDa). CE-SDS revealed pronounced LMW increases under stressed conditions; mass spectrometry mapped proteolytic cleavage sites. A protease inhibitor panel identified pepstatin A as selectively suppressing cleavage, implicating residual aspartic protease activity and guiding process improvements to enhance product stability.

Co-formulation of therapeutic proteins can reduce dosing errors, simplify administration, and improve patient compliance, but it also increases analytical complexity for attribute characterization and stability assessment. Here, we established an integrated analytical strategy for co-formulated monoclonal antibodies (mAbs) that includes size-exclusion UHPLC (SE-UHPLC) to monitor high-molecular-weight (HMW) species, cation-exchange HPLC (CEX-HPLC) to monitor charge variants, and LC-MS/MS peptide mapping for protein-ratio determination and critical quality attribute (CQA) monitoring.

Adeno-associated virus (AAV) vector aggregation is a key concern in gene therapy because it can compromise product efficacy and safety. This presentation examines AAV aggregation driven by thermal, mechanical, and freeze-thaw stresses; by formulation factors including pH, ionic strength, and excipients; and by impurities. The impact of aggregation on in vitro biological activity is also evaluated. These findings provide insight into aggregation mitigation and support formulation development and manufacturing investigations.

This work demonstrates that Arrhenius-based accelerated predictive stability (APS) can reliably predict siRNA shelf life. APS complements conventional stability studies, with predictions closely matching real-time data. This enables conservative, data-driven shelf-life assessments, supports informed long-term stability decisions, and accelerates development timelines to help siRNA therapies reach patients faster.

To identify potential developability issues early in the discovery process, the Biologics Research Center at Novartis uses in-house data to train physics-based and machine learning models that predict key attributes including PTMs, hydrophobicity, and thermal stability from molecular sequences or structures. This talk will highlight recent progress and show how integrating these tools into discovery workflows can improve pipeline quality and reduce costly downstream issues.

Draft ICH Q1 guideline extends accelerated stability testing to biopharmaceuticals. Temperature represents the most common acceleratory factor, and Arrhenius equation is commonly used to analyze temperature dependence of degradation rate constants. While such approach has been successfully applied to various drugs, there are cases of significant deviations from Arrhenius behavior. Scientific mechanisms for non-Arrhenius temperature dependence of degradation are discussed, and refined criteria for selecting accelerated storage conditions are proposed.

Early biologics discovery requires predictive approaches that connect molecular sequence with structural context and measurable biophysical behavior. By integrating sequence liabilities, three-dimensional descriptors, residue-level physicochemical features, and experimental developability data, interpretable machine-learning models can identify risk drivers before extensive characterization. This unified framework enables earlier triage of antibodies and VHHs, supports rational engineering of problematic regions, and informs library design strategies that improve manufacturability, stability, and candidate selection.

Advances in LC-MS proteomics now enable large-scale biological measurements, but interpreting these complex datasets remains a major bottleneck. We present an end-to-end platform that combines scalable proteomic workflows with agentic AI to rapidly transform high-dimensional proteomic data into mechanistic insights and testable biological hypotheses. By integrating automated data analysis, pathway interpretation, literature evidence, and hypothesis generation, our framework reduces analysis time from days to hours while improving biological interpretability. We will demonstrate its application to large-scale drug perturbation studies and discuss how AI-enabled workflows can accelerate target discovery, mechanism-of-action studies, and translational proteomics.

Cryo-electron microscopy (cryo-EM) resolves antibody-antigen complexes andconformational states at near-atomic detail, but conventional throughputcannot keep pace with biologics engineering cycles. This presentationdescribes an integrated pipeline coupling high-throughput cryo-EM withmachine-learning-driven image processing and structural analysis to movefrom sample to decision at campaign speed. Rapid, iterative structures feedstructure-enabled ML: epitope and paratope mapping to guide interface design,conformational analysis to flag developability liabilities, and experimentallygrounded structures to train generative-design and affinity models. Positionedthis way, cryo-EM becomes a first-class signal in an active-learning looprather than a terminal validation step, accelerating engineering decisions.

Protein–protein interactions affect antibody developability (solubility, viscosity, stability, aggregation), but bispecifics remain poorly understood. We built a computational-experimental framework using coarse-grained simulations to predict B{22} for ~102 monoclonal antibodies, then generated a virtual bispecific library by combining variable regions. Selected candidates were characterized via light scattering, confirming condition-dependent interactions and validating the framework's predictive power.

Structure-based hydrogen-deuterium exchange mass spectrometry (HDX-MS) experiments were combined with in silico analysis to design IgG1 variants with improved developability properties. Resulting variants had improved viscosity at high concentration and/or reduced subvisible particles while maintaining target binding affinity. Analysis using machine learning (ML) approaches incorporating experimental data and computed descriptors revealed molecular determinants of subvisible particle formation, providing a framework for risk mitigation of late-stage antibody drug development attributes.

Regeneron is experiencing an increase in the number and diversity of novel alternative protein formats entering the development pipeline. Therefore, it became essential to adapt first-in-human Chemistry, Manufacturing, and Control (CMC) workflows developed for monoclonal antibodies to address novel challenges associated with these formats. An Alternative Format Workshop Team was created to address these cross functional opportunities. Central to this work, is (i) a cross-functional, prospective evaluation of product-related impurity risks, and (ii) formalizing knowledge transfer between stakeholders in research, preclinical, and CMC functions. The team enhanced workflows by introducing stage appropriate risk assessments and formalizing cross functional decision points and work allocation. This provided a robust data package to inform product quality goals for process development as well as release specifications. This presentation will highlight the drivers behind the formation of the Alternative Format Workshop Team, provide additional details on the workflow enhancements implemented, and illustrate these improvements through a case study.