Search papers, labs, and topics across Lattice.
TriFit, a novel multimodal framework, integrates protein sequence, structure, and dynamics to predict the functional impact of single amino acid substitutions (SAVs). It uses a four-expert Mixture-of-Experts (MoE) fusion module with trimodal cross-modal contrastive learning to combine embeddings from ESM-2, AlphaFold2, and Gaussian Network Model (GNM). Results on the ProteinGym benchmark demonstrate that TriFit outperforms existing supervised and zero-shot baselines, with protein dynamics providing the largest marginal contribution.
Protein dynamics, often overlooked, are the secret sauce for predicting the impact of mutations, boosting accuracy beyond sequence and structure alone.
Predicting the functional impact of single amino acid substitutions (SAVs) is central to understanding genetic disease and engineering therapeutic proteins. While protein language models and structure-based methods have achieved strong performance on this task, they systematically neglect protein dynamics; residue flexibility, correlated motions, and allosteric coupling are well-established determinants of mutational tolerance in structural biology, yet have not been incorporated into supervised variant effect predictors. We present TriFit, a multimodal framework that integrates sequence, structure, and protein dynamics through a four-expert Mixture-of-Experts (MoE) fusion module with trimodal cross-modal contrastive learning. Sequence embeddings are extracted via masked marginal scoring with ESM-2 (650M); structural embeddings from AlphaFold2-predicted C-alpha geometries; and dynamics embeddings from Gaussian Network Model (GNM) B-factors, mode shapes, and residue-residue cross-correlations. The MoE router adaptively weights modality combinations conditioned on the input, enabling protein-specific fusion without fixed modality assumptions. On the ProteinGym substitution benchmark (217 DMS assays, 696k SAVs), TriFit achieves AUROC 0.897 +/- 0.0002, outperforming all supervised baselines including Kermut (0.864) and ProteinNPT (0.844), and the best zero-shot model ESM3 (0.769). Ablation studies confirm that dynamics provides the largest marginal contribution over pairwise modality combinations, and TriFit achieves well-calibrated probabilistic outputs (ECE = 0.044) without post-hoc correction.
