A framework for the rational selection of a minimal suite of non-degenerate developability assays (DAs) that maximise insight into candidate developability or storage stability is lacking. To address this, we have subjected a panel of test mAbs to a range of distinct DAs, and also assessed their long-term storage stability. We show that it is possible to identify a reduced set of key variables from this suite of DAs using orthogonal statistical methods.

Using insights from the crystal structure of anti-Hu/Cy CD3 antibody ADI-26906 in complex with CD3 epsilon (CD3e) and antibody engineering using a yeast-based platform, we have derived high-affinity CD3 antibody variants with very low polyreactivity and significantly improved biophysical developability. Comparison of these variants with CD3 antibodies in the clinic (as part of bi- or multi-specifics) shows that affinity for CD3e is correlated with polyreactivity. Our engineered CD3 antibodies break this correlation, forming a broad affinity range with no to low polyreactivity.

Intrinsic biophysical properties can impact the pharmacokinetics of candidate therapeutic mAbs. We developed and embedded a high-throughput in vitro screen to test in vivo suitability of lead panels of candidate antibodies and this screen is now a critical piece of our new dose prediction strategy.

Much progress has been made for the (developability) property prediction of antibodies using AI/ML methods, allowing the design of huge sets of sequences in silico. While these approaches are feasible for standard monospecific antibodies, they are often not applicable for more complex next-generation antibodies (including multispecifics and ADCs). This presentation will showcase lessons learned and specific applications of physico-chemical property prediction strategies to assess and even optimise bispecifics and ADCs.

The flexible BEAT platform enables 5 or more functional modules to be combined into a single molecule. The biophysical properties of a complex multi-specific immune cell engager antibody can be quite different to the sum of its parts. Therefore, a developability screening cascade was developed starting from Fab or cytokine selection to multi-specific lead candidate selection. This was applied to identify ISB 2001, a first-in-class tri-specific BCMA and CD38 T cell engager now advancing in the clinic to treat Multiple Myeloma.

Liability evaluation based on the 3D structures of biotherapeutics alone or in complex with their target provides critical insights. When experimental structural data are unavailable, computational modeling can aid in generating plausible structures. By combining these models with functional data, we can assess the criticality of the liabilities. Additionally, docking methods and other computational strategies can be employed for interface redesign, potentially enhancing stability and thereby improving development outcomes.

• Native TCRs suffer from poor intrinsic stability and low expression yields when produced as soluble proteins. 
• We developed a high-throughput TCR yeast surface display system and applied it to screen TCR mutagenesis libraries by thermal cycling followed by assessment of retained antigen binding. 
• We coupled this approach with deep sequencing and computational analysis of the resulting datasets to identify stabilised TCR variants with enhanced biophysical properties. 
• We identified a minimal set of universal mutations that confer soluble TCRs with enhanced thermal stability, aggregation resistance and expression yield, as well as with a low potential for immunogenicity.?

With the rise of AI, Biologics drug discovery is evolving, opening up new possibilities for developability assessment. Our InSiDe (in silico developability) pipeline enables scoring of antibodies and nanobodies, facilitating the selection of more developable leads. We will introduce our "lab in the loop" strategy for developability, built on machine learning, high-throughput assays, and IT integration.

BigHat Biosciences has developed novel machine learning (ML) approaches that leverage our high speed, automated wet lab in order to rapidly and iteratively design over a thousand next generation therapeutic antibodies each week. Our algorithmic approach pairs with our automated wet lab to guide the search for better molecules by learning from each cycle of characterization across affinity, function, and developability measures of each antibody. We’ll highlight key features of our platform and computational methods as well share several case studies of novel next-generation antibodies designed by solving increasingly challenging multi-parameter optimization challenges on our platform. These applications will focus on how we can systematically design safe and effective TCEs for classic solid tumor targets and jointly optimize many properties to create next generation ADCs.