CIS Seminar: “Probabilistic Experimental Design for Petascale DNA Synthesis”
March 18 at 3:30 PM - 4:30 PM
Generative modeling offers a powerful paradigm for designing novel functional DNA, RNA and protein sequences. In this talk, I introduce probabilistic experimental design methods to efficiently manufacture samples from generative models of biomolecules in the real world. These algorithms merge computational techniques for approximate sampling with physical randomness. I also develop tools to rigorously evaluate the quality of manufactured samples, including nonparametric
two-sample tests with consistency guarantees and scalable algorithms. I demonstrate synthesizing ~10^16 samples from a generative model of human antibodies, at a sample quality comparable to state-of-the-art protein language models, and a cost of ~$10^3. The library yields candidate therapeutics for “undruggable” cancer targets. Using previous methods, manufacturing a DNA library of the same size and quality would cost roughly ~$10^15.
Eli N. Weinstein
Data Science Institute, Columbia University
Eli N. Weinstein is a postdoctoral research scientist with David Blei at Columbia University, in the Data Science Institute. He also serves part-time as Director of Machine Learning Research at Jura Bio, a genomic medicines startup. His research is in probabilistic machine learning and its application to the molecular sciences. Previously, he received his PhD in Biophysics from Harvard University in 2022, advised by Debora Marks and Jeffrey Miller, and an A.B. in Chemistry and Physics with highest honors from Harvard in 2016, advised by Adam Cohen. He is the recipient of various awards including best paper awards at the Molecular Machine Learning Conference and the New England Statistics Symposium, an oral presentation at Neural Information Processing Systems (NeurIPS), and a Hertz Foundation Fellowship.