Antibody-based therapeutics enjoy considerable successes as cancer treatments, but can cause serious toxicities due to recognition of tumor-associated antigens in non-cancerous tissues. We will discuss recent efforts to develop advanced antibody therapeutics with Fc-mediated activities that are restricted to the acidic solid tumor microenvironment. With the intent of decreasing toxicities and expanding therapeutic windows, protein engineering strategies can render antibody activity sensitive to multiple tumor-specific characteristics.

The antibody Fc domain shapes protective immunity against many different viral pathogens. We will present a Systems Serology approach to define antibody Fc features associated with protection from Ebola virus and describe how we have used antibody engineering approach to validate and mechanistically dissect the role of distinct Fc-mediated functions in protection. Further, we will discuss how to translate these approaches to other pathogens, such as SARS-CoV-2.

In vitro biopanning platforms have expanded the field of antibody identification beyond immunization. However, applying these strategies to identifying binders against challenging targets remains a challenge. Here, we present a new pipeline, RAPID (Rare Antibody Phage Isolation and Discrimination), for the identification of rare high-affinity antibodies against challenging targets. RAPID is enabled by fluorescent labeled phage displayed antibody libraries, and fluorescent activated sorting allows for the isolation of promising binder populations. A Biolayer Interferometry method allows for candidate clones to be discriminatorily screened. We introduce this method with a well characterized Antigen-Antibody pair, and applications of RAPID biopanning are presented.

Antagonist antibodies for membrane protein targets like GPCRs and ion channels are difficult to discover. Agonist and modulating antibodies are even more challenging. Here I will discuss the challenges in the field and strategies for discovering and developing antibodies that access the full spectrum of activities against membrane protein targets.

Molecular simulation techniques can help solve the inverse problem of what antibody structures give rise to an experimental observable and provide valuable insight into the biophysical origin of antibody properties. Here, we use coarse-grain and all-atom molecular dynamics to accurately model how far antibodies can reach to bind antigens at both arms and to reveal differences in how TCR-mimetic antibodies bind pMHC targets in comparison to TCRs.

A significant barrier to cancer treatment is that although small molecule drugs easily enter cells, many intracellular oncoproteins lack suitable binding pockets, and so are unresponsive to them. Antibodies can act across a broad surface, so inhibiting protein-protein interactions, but unfortunately do not naturally enter cells. We have developed multimeric, cyclised cell-penetrating peptides that transfer functional antibodies efficiently across the cell membrane, allowing the targeting of previously “undruggable” intracellular molecules.

For the development of safe and effective T Cell Engagers (TCEs) for solid tumors, it is highly desirable to expand the number of available target antigens and also to increase the precision with which a TCE can differentiate tumors from healthy tissue. I will present a strategy to reduce on-target, off-tumor TCE activity by targeting co-expression of two tumor-associated antigens and describe engineering of trispecific antibodies with this selectivity.

• Current successes
• Experimental validation and POC
• Bottlenecks and challenges
• Needs from IT and solution providers

• What preclinical in vivo data are most often predictive of clinical success
• What are some ways we could increase in vivo throughput
• How can AI and machine learning be utilized in conjunction with in vivo data
• How to improve the in vitro to in vivo drug development process by supplementing with early in vivo discovery workflows
• Unbiased in vivo based therapeutic discovery and the required in vivo throughput

In the eternal race to make better biotherapeutics faster, GSK Biopharm discovery pipelines are continuously revamped by a tight collaboration among different departments. This presentation will focus on the computational methods that help select better targets, design and enrich screening libraries, and accelerate the multiobjective optimization of lead candidates through the analysis of NGS data with developability criteria.

When engineering therapeutics, in vivo validation remains the primary bottleneck to program advancement. We invented a novel method that enables multiplexed quantification of 100s of protein therapeutics in vivo. We have leveraged our in vivo data with machine learning to perform in vivo molecule design. Further, we show that in vitro assay results are often poorly predictive of downstream in vivo results, highlighting the importance of increased in vivo throughput.

Effective navigation and distillation of large structural datasets necessitates programmatic approaches. We describe the custom structural informatics platform (GYST) we are developing and its application to study thousands of fab interfaces for opportunities to engineer novel polyvalent fab formats. This platform allows expansion to other protein families within and beyond large molecule drug discovery. We also discuss how the amassed, pre-computed annotations are suited for training deep learning models.

VHHs or nanobodies are single antigen binding domains originating from camelid heavy-chain antibodies. Analyzing variable domain structures from llama and alpaca we found that VHHs can be classified into structural clusters based on their CDR-H3 conformation. The clusters have distinct functional properties in how they interact with the antigen. VHHs from the two clusters originate from different VH germlines, demonstrating a previously undescribed impact of germline usage on CDR3 conformation.

Therapeutic antibodies have played a critical role in healthcare for over three decades, yet the rate of discovering novel antibodies remains outpaced by the need for treatment options. This is particularly true for challenging glycan targets such as tumor-associated carbohydrate antigens (TACAs). In this work, we use Rosetta software to optimize affinity and specificity among a large collection of natively-paired, immune-evolved antibodies.