Portal is the home of the AI for drug discovery community. Join for more details on this talk and to connect with the speakers: https://portal.valencelabs.com/logg

Note: This session was zoombombed. We tried to edit the video but there may be some interruptions or imagery.

Abstract: We resolve difficulties in training and sampling from a discrete generative model by learning a smoothed energy function, sampling from the smoothed data manifold with Langevin Markov chain Monte Carlo (MCMC), and projecting back to the true data manifold with onestep denoising. Our Discrete WalkJump Sampling formalism combines the maximum likelihood training of an energybased model and improved sample quality of a scorebased model, while simplifying training and sampling by requiring only a single noise level. We evaluate the robustness of our approach on generative modeling of antibody proteins and introduce the distributional conformity score to benchmark protein generative models. By optimizing and sampling from our models for the proposed distributional conformity score, 97100% of generated samples are successfully expressed and purified and 35% of functional designs show equal or improved binding affinity compared to known functional antibodies on the first attempt in a single round of laboratory experiments. We also report the first demonstration of longrun fastmixing MCMC chains where diverse antibody protein classes are visited in a single MCMC chain.

Speaker: Nathan Frey   / nc_frey  

Twitter Hannes:   / hannesstaerk  
Twitter Dominique:   / dom_beaini  

~

Chapters
00:00 Intro + Background
03:35 Learning Scores on a Smooth Manifold
16:00 Discrete WalkJump Sampling
25:58 WJS Produces High Fitness Molecules
38:10 Discussion
54:50 Conclusion