Selection of lead antibodies requires evaluation of large numbers of molecules to find those with optimal specificity, stability, and predicted post-translational modifications and immunogenicity, as well as affinity, epitope, and biological function. Here, we describe our high-throughput workflow for cloning and production of hundreds of antibodies at the 1 mg scale and how it is integrated with multiparameter evaluation to enable rapid, consistent decision-making. Applications to bispecifics will be discussed.

Deep learning, as part of a family of tools related to machine learning, is an emerging field of information and computer science that uses large data to identify complex relationships. Here, I will describe how we are moving beyond experimental screening by applying deep learning to augment therapeutic antibody optimization in mammalian cells.

Antidrug antibodies can undermine biotherapeutic efficacy and threaten patient safety, and mitigating immunogenicity risk is therefore of increasing importance in biopharma. To date, even the most proactive mitigation strategies occur as part of final lead selection or alternatively re-engineering of a lead candidate. Here, we describe immunogenicity mitigation as part of a design basis approach to the earliest stage of biotherapeutic discovery: combinatorial library construction and screening.

The risk assessment approaches have evolved over the past decade with multiple approaches that rely on both algorithms and cell-based assays. This talk will provide recent advances in several cell-based assay formats to address specific questions around immune modulation, next-generation biologics, and gene therapy/cell-based and nucleic acid-based therapies. The questions around antigen processing and the T cell repertoires to react with the biologics will be presented.

Recognizing the growing body of literature around the impact of developability on therapeutic antibody development, our group has implemented a workflow for developability screening of candidates at various stages of antibody discovery and optimization. We will present our current workflow, along with several case studies which demonstrate the role of developability assessment as a risk mitigation tool in antibody discovery.

The development of microfluidic platforms has allowed us to dramatically reduce the timelines needed to identify novel therapeutic antibodies. These devices allow us to identify binder panels within minutes and more complex screens can be completed within hours. In this talk, I will discuss the challenges encountered in developing an antibody discovery platform on the Berkeley Lights BeaconTM and the strategies that we used to overcome the obstacles.

Biologics pipelines are focusing on increasingly sophisticated protein formats, supplementing traditional monoclonal antibodies with bispecifics, fusion proteins, and other biological entities. These new formats can lack the stability required for therapeutics, so generalized methods are required to engineer sufficient stability without compromising biological function. I will describe our implementation of computational modeling techniques, guided by biophysical characterization, to enhance protein stability in our high-throughput protein engineering workflows.

The selection of antibody libraries displayed on phage against live cells is notoriously challenging due to the stickiness of bald phage. A new method, termed Fab biotinylation and capture (FBC), was developed to address this shortcoming. Following proof-of-concept, FBC was used for differential selection on target cells versus non-target cells followed by next-generation sequencing (NGS). Combining FBC and NGS was applied to antigen discovery in multiple myeloma.

Traditional antibody discovery approaches incorporate a screening funnel with medium- to high-throughput screening for binding, leading to low-throughput screening of leads in functional assays. We developed automated approaches, enabling us to incorporate high-throughput functional screening of thousands of antibodies and other molecules in the final therapeutic format in the primary screening phase. This enables rapid identification of functional epitopes at the beginning of the discovery process.

The traditional route to antibody discovery and repertoire profiling is by proxy of the B-cell repertoire localized within the secondary lymphoid organs or transiently circulating within the blood. Here, we demonstrate the utility of high-resolution mass spectrometry to directly identify serum IgG antibody with molecular-level resolution (Ig-seq) in a variety of disease settings, including COVID-19. Further, we have recently extended this approach to the analysis of purified circulating immune complexes (CICs), allowing tandem discovery of both serum antibodies and the antigens to which they bind. Ig-seq of CICs holds particularly strong potential in cases where the identity of the antigen is unclear. Moreover, bloodborne CICs have been universally detected in all known viral diseases of man, and in some instances have been correlated with disease activity and severity, but have never been examined in detail nor capitalized upon for comprehensive analysis of human humoral immune responses. We shall apply Ig-seq proteomics for the simultaneous molecular decomposition of polyclonal antibody repertoires and the viral antigens they target.

AbCellera is a leader in antibody discovery and has established rapid and robust responses to viral outbreaks with its Pandemic Prevention Platform (P3). The company’s core technology is a high-throughput microfluidic platform that uses miniaturized assays to screen antibodies from millions of individual immune cells from any species, including directly from human donors. By combining its core technology with Celium™, the world’s first data mining and antibody visualization software, AbCellera’s platform can search and analyze natural immune systems faster and deeper than other technologies to deliver exponentially more antibodies against any disease. 

On receiving a sample from one of the first US patients to recover from COVID-19, AbCellera screened around 6 million immune cells and in just 3 days, discovered more than 2,000 antibodies that bind to the SARS-CoV-2 spike protein, which is used by the virus to attach to and infect cells. From these, over 500 unique human antibodies were sequenced and identified in under 5 days. Using our proprietary bioinformatics visualization software, Celium, these were down-selected to a set of potential therapeutic candidates, which were cloned, produced and characterized, and sent to the Vaccine Research Centre (VRC) for validation testing. Celium was used to further down-select to the most promising antibody leads. These were delivered to AbCellera’s partner, Eli Lilly and Company, for drug development and clinical trial testing. 

In just 23 days, AbCellera identified lead candidate antibodies that could be deployed as a treatment or to prevent SARS-CoV-2 infection.