Siqi Sun

FoldBench evaluates all-atom biomolecular structure prediction models across nine tasks, revealing performance gaps and highlighting AlphaFold 3's superiority.

The paper introduces BALM, a bio-inspired antibody language model, and BALMFold, a method for predicting antibody structures, outperforming existing models in benchmark tests.

A BERT-style model pretrained on 23.7 million unlabelled ncRNA sequences gives embeddings that improve secondary structure, contact, distance-map and some function predictions over task-specific and MSA-based baselines.

fastMSA accelerates multiple sequence alignment by using a dual encoder based on a protein language model, achieving a 34-fold speed-up while maintaining homolog recall and compatibility with downstream tasks like AlphaFold.

The paper evaluates a deep learning method for protein contact prediction in CASP12, achieving 47% and 44% accuracy for long- and medium-range contacts, respectively.

This paper introduces a deep learning method for predicting protein contact maps, achieving higher accuracy than existing methods, especially for proteins with few sequence homologs.