Event
Abstract: Generative models offer a powerful paradigm for designing novel functional DNA, RNA and protein sequences. In this talk, I introduce a method to efficiently synthesize designs from generative models in the real world. The method involves a combined machine learning and wet lab procedure that implements generative sampling algorithms physically, through controlled stochastic chemical reactions. I demonstrate synthesizing ~10^17 designs from a generative model of human antibodies, at a level of realism and diversity comparable to state-of-the-art protein language models, and a cost of ~$10^3. The library yields therapeutic candidates against HLA-presented intracellular tumor antigens. Using previous methods, a library of the same size and quality would cost roughly ~$10^15.