Computational design of synthetic antibody repertoires for accelerated therapeutic discovery

Introduction

Synthetic human antibody repertoires are an important source of therapeutics; however, antibodies often exhibit undesirable developability liabilities, such as low stability, solubility, and polyreactivity, that limit their potential as drug candidates. CADABRE will design next-generation repertoires comprising billions of diverse human antibodies that exhibit excellent developability properties, with the following aims:

Repertoire Design

1. Combinatorial Design: We will use our newly developed combinatorial Rosetta atomistic design paradigm to select human germline genes and design H3 multipoint mutants.
2. Diversity Creation: These will combine into billions of diverse, low-energy, and foldable full-length antibody variable domains for expression in a phage-displayed repertoire.
3. Structural Diversity: The repertoires will comprise hundreds of possible germline gene combinations, increasing the structural diversity relative to existing repertoires and the odds of obtaining diverse antibodies toward any antigen.

Learning Developability Principles

• High-Throughput Screening: High-throughput screening will identify heat-stable antibodies that are not polyreactive.
• AI-Based Predictor: Data from deep sequencing will be used to train an AI-based predictor of these properties that will be used to improve the repertoire and rank antibody candidates.
• Iterative Process: We will iterate repertoire design, screening, and learning until we converge on a repertoire that exhibits excellent properties.

Verifying Relevance to Therapeutic Discovery

• Target Selection: We will select antibodies that target antigens that represent relevant drug targets.
• Property Verification: This will verify that the antibodies exhibit high stability, affinity, and developability.

Conclusion

CADABRE combines the strengths of our protein-design methods (generating stable antibodies) and phage-display screening (unbiased binder selection). It will deepen our understanding of the biophysical underpinnings of antibody developability, develop new methods for ranking drug candidates, and generate new repertoires that will accelerate the discovery of life-saving therapeutic antibodies.