Dianzhuo Wang

The paper introduces BioSecBench-Function, a benchmark for assessing AI agents' ability to infer biological function from experimental data, showing varying performance across models and tasks.

The paper investigates how metabolon clustering enhances coenzyme Q biosynthesis through substrate channeling using simulations and experiments.

The paper develops a probabilistic framework to predict viral immune escape trajectories, showing that escape is constrained to specific pathways despite vast mutational space.

This paper introduces a framework that integrates biophysical priors into protein embeddings to improve DeltaDeltaG prediction, showing that general-purpose encoders can outperform specialized models.

This paper introduces a few-shot learning framework for detecting high-fitness viral variants using Bayesian active learning and biophysics, showing it can identify variants up to fivefold faster with fewer experiments.

The paper introduces PROXICLUST, a method that uses spatial clustering of interface residues to predict viral protein binding affinities across larger mutation spaces using smaller datasets.

This study shows that using consistent structures improves ESM3's binding prediction, but using different relaxed mutant structures reduces performance.

This study identifies entropy changes in water networks as a mechanism for ion-induced protein denaturation, revealing an 'inverse hydrophobic effect' that promotes unfolding.