Zishuo Zeng

This study benchmarks machine learning models for enzyme engineering using a curated dataset, finding zero-shot predictors are most generalizable but still modestly correlated with activity, with consensus models improving precision.

This paper introduces a framework combining a protein GPT model with a custom discriminator to prioritize diverse and functional protein sequences for de novo design.

The paper evaluates splicing prediction methods on 56 clinically validated variants, finding some methods match or exceed SpliceAI's performance but highlighting ongoing challenges in accuracy.

The paper introduces synVep, a machine learning method for predicting the effects of synonymous variants, achieving high precision and recall while improving differentiation between pathogenic and benign variants.