Chris J. Maddison

A reasoning LLM combining ESM3 embeddings with GO-GPT term predictions produces protein function annotations that experts rated tie-or-better than UniProt in 79% of cases (SFT model).

This study shows that fine-tuning protein models on text annotations improves their predictive power for function prediction tasks, outperforming BLAST in some cases.

The paper introduces kNN-AS, an adaptive sampling method that uses k-nearest neighbour graphs to efficiently explore conformational space, showing state-of-the-art performance on artificial and protein systems.