The broader impact of this Small Business Innovation Research (SBIR) Phase I project is to improve and accelerate the development of antibody therapies to fight the COVID-19 pandemic. Antibody therapies are in development to target SARS-CoV-2, the virus responsible for COVID-19, but emerging virus mutations may result in resistance and require the development of new therapies. Antibody therapies typically enter the clinic after a year of development or longer, and the medical and economic consequences of this delay are severely felt throughout the world. To reduce the time to produce an antibody against an emerging SARS-CoV-2 mutant, this project develops a computational platform trained using massive experimental datasets to rapidly predict the therapeutic potency. This platform will enable drugs against SARS-CoV-2 mutants to more rapidly reach the clinic, saving thousands of lives. Moreover, the proposed platform can be utilized to predict therapeutic efficacy against future coronavirus strains unassociated with the COVID-19 pandemic, providing an invaluable tool to fight future pandemics.The proposed project will demonstrate the feasibility of using quantitative and library-on-library protein interaction datasets to train machine learning models for predicting antibody binding to novel SARS-CoV-2 variants. Existing approaches to build computational predictions for antibody drug development have been limited to few target variants, since datasets with binding measurements against hundreds or thousands of targets are not available. This project involves optimizing and validating a cell-based platform for generating a sufficient quantity and quality of antibody-antigen binding data for training computational models. The platform uses genetically engineered yeast cells and next generation sequencing to link protein interaction strength with cellular mating frequency. To demonstrate feasibility, large multi-chain antibody libraries will be genomically integrated in yeast and enriched for binding to SARS-CoV-2 and related coronaviruses. Next, a large network of antibody-antigen interactions will be measured and validated for quantitative accuracy by comparing to biophysical measurements. Finally, the resulting data will be used to train machine learning models and evaluate their predictive power using cross-validation. Training of computational models with sufficient predictive power will demonstrate the feasibility of using quantitative and library-on-library binding data coupled with machine learning to develop a platform for rapid antibody development to a novel SARS-CoV-2 mutant.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
