Virtual Seminar Series on Gaussian Processes, Spatiotemporal Modeling, and Decision-making Systems

2023-04-24

Treating rapidly evolving pathogenic diseases, like COVID-19, requires a therapeutic approach that tolerates the emergence of new viral variants over time. Our machine learning (ML)-guided sequence design platform combines high-throughput experiments with machine learning to generate highly diverse single-domain antibodies (VHHs) that tightly bind and neutralize SARS-CoV-1 and SARS-CoV-2. Crucially, the model trained using binding data against multiple SARS-CoV variants accurately captures the relationship between VHH sequence and activity across a broad swathe of VHH sequence space, resulting in ML-designed VHHs that neutralize targets not seen during training, including the Delta and Omicron variants of SARS-CoV-2, both of which escape neutralization from the parent antibody. We found numerous ML-designed VHHs with significantly improved activity 5-15 mutations away from their parent sequences.

Zelda is a senior research scientist within Google Brain. She did her PhD at MIT, working with Suvrit Sra and as a member of the Machine Learning and Learning and Intelligent Systems groups. Her research focuses on identifying precise mathematical definitions of diversity to understand the behavior of ML models, e.g., under distribution shift. Zelda did her PhD on negatively dependent measures, which use Strongly Rayleigh polynomials to encode desirable properties for diversity modeling.