Recent & Upcoming Talks

Substrate-aware zero-shot predictors can accelerate enzyme engineering by estimating activity for non-native substrates and new-to-nature reactions, with a weighted ensemble of AlphaFold 3 and EVmutation scores demonstrating broad generalization across diverse chemistries.

Co-organize The Generative and Experimental Perspectives for Biomolecular Design (GEM) workshop at ICLR 2025 to foster collaboration between machine learning experts and experimental scientists.

We conducted the most comprehensive transfer learning study for protein language models to date and found that scaling pretraining does NOT scale downstream task performance, except for structure prediction

Co-organize The Generative and Experimental Perspectives for Biomolecular Design (GEM) workshop at ICLR 2024 to foster collaboration between machine learning experts and experimental scientists.

A poster presentation on multi-modal representation learning for predicting top protein fitness from combinatorial and random mutagenesis libraries

A video abstract on Pretrained protein language model transfer learning: is the final layer representation what we want?