Ariel Tennenhouse is a Ph.D. student with Sarel Fleishman at the Weizmann Institute of Science. He did his bachelor’s degree in chemical biology at UC Berkeley and moved to the Weizmann Institute in 2019. He is a member of the Teva National BioInnovators Forum and the Azrieli Fellows Program. His research focuses on developing one-shot computational methods to co-optimize antibodies for many properties that are critical to therapeutic development. He is now applying these methods to design synthetic antibody repertoires for accelerated discovery of developable antibodies. Broadly speaking, Ariel is excited about using protein engineering and design to generate large-scale datasets to learn fundamental principles about proteins and biological functions.

Monoclonal antibodies have revolutionized clinical care of a wide range of previously untreatable diseases. However, developing antibodies into therapeutics remains challenging. One reason for this is that to be useful therapeutics, antibodies must excel in multiple characteristics (such as affinity, stability, and specificity) and engineering an antibody to excel in all necessary criteria is challenging. Most antibody engineering pipelines implement experimental procedures based on random or semi-random processes that tend to trade off gains in one desired characteristic with losses in others.

We are developing two complementary methods combining atomistic design and machine learning for one-shot antibody optimization. Our work demonstrates that designing the antibody variable fragment for stability can improve antibody affinity and developability without prior experimental data and with twenty or fewer designs tested. We can additionally generate libraries containing thousands of low-energy designs that can be screened for further improvements and used as training data for machine learning. We are also developing an approach for atomistic design of synthetic antibody repertoires based on hundreds of antibody frameworks for accelerated discovery of developable antibodies. We hope this will lead to a platform from which developable antibodies can be directly selected, reducing the need for optimization of individual antibodies.

• We’ve developed two complementary, physics-based methods that can reliably optimize antibodies for a variety of characteristics.
• We are now applying these methods to design universal synthetic antibody repertoires in which every variant is designed for developability.
• We’ve constructed a proof-of-concept repertoire and identified specific binders against four diverse antigens.