2026/05/21 by Taewon Kim, Hyosoon Jang, Hyunjin Seo +6 · 1 voice
Biochemistry, Genetics and Molecular Biology · Computer Science · Materials Science · #Benchmark (surveying) #Enzyme Structure and Function #Function (biology) #Generative grammar #Generative model #Machine Learning in Materials Science #Protein Structure and Dynamics #Representation (politics) #Set (abstract data type) #Training set #cs.AI #q-bio.BM
paper · pdf · open access · doi:10.48550/arxiv.2605.22133
published in arXiv (Cornell University) (Cornell University)
openalex publication_date 2026/05/21 · arxiv published 2026/05/21 · openalex created_date 2026/05/23 · arxiv updated 2026/05/26 · openalex updated_date 2026/07/28
Recent advances in generative modeling show that pretrained representations can improve generation as conditioning features or alignment targets. Motivated by this, we study protein representations for predicting structures beyond conventional function annotation. We propose TriProRep, a structure-aware pretraining method that jointly models three aligned residue-level views: amino-acid identity, backbone geometry, and local full-atom geometry, discretely encoded via VQ-VAE tokenizers. By pretraining to recover original tokens from generator-corrupted views, TriProRep learns to distinguish plausible but incorrect cross-view augmentations from the original protein. We further introduce RepSP, a benchmark for evaluating protein representations in structure-predictive settings. RepSP tests three uses of representations: homodimer co-folding from apo-chain representations, residue-level prediction of homodimer-derived interaction properties, and representation-aligned monomer structure prediction. Across these tasks, TriProRep improves over sequence-only and prior structure-aware representation models, while maintaining competitive performance on conventional benchmarks.