CHRISTOPHER JAMES LANGMEAD

This paper shows that protein language models can better annotate metagenomic proteins when trained on curated data, enabling retrieval of diverse enzyme candidates from billions of sequences.

This study analyzes how biological concepts are encoded in protein language models across their layers and finds that complexity increases with depth, with pretraining data and compute having a greater impact than parameter scaling.

This paper evolves computationally designed protein binders using in vivo hypermutation and maps their sequence-affinity landscapes to understand affinity improvements and mutation interactions.

This paper introduces AMPLIFY, a protein language model that achieves better performance with less training cost by improving data quality and architecture, challenging the need for scaling.

HARVESTMAN is a framework for hierarchical feature learning and selection from whole genome sequencing data, showing improved performance over binary SNP representations in breast cancer classification.