Advancing science and technology in the biopharmaceutical industry.

2020-04-13  bioRxiv

Designing Feature-Controlled Humanoid Antibody Discovery Libraries Using Generative Adversarial Networks

Just Evotec Biologics has compiled a library of a new class of humanoid antibodies using a novel, AI-derived approach. Antibodies are proteins produced and secreted by B-cells. As they bind to foreign substances that invade the body, such as pathogens, they represent an important class of biologics-based therapeutics. However, patient accessibility is currently quite limited, as B-cells do not undergo selective pressure in vivo to produce antibodies with ideal biotherapeutic characteristics. Additionally, protein scientists have traditionally had to rely on the very limited scope of previously-discovered antibodies and their close variants to develop predictive models of complex antibody behavior.

Just – Evotec Biologics’ data and laboratory scientists have developed a novel antibody library generation method using Generative Adversarial Networks (GANs), that enables controlled design of biological and structural features for therapeutic quality and developability. This approach creates a new class of AI-derived humanoid antibodies that mimic the diversity and properties of the human antibody repertoire while optimizing the antibodies for development, manufacturing, and in use characteristics that make them better suited as biotherapeutics. HAL is expected to serve as a highly efficient starting point for drug discovery and development projects across a broad range of indications.   link to bioRxiv