Enhancing Clinical Classification of Protein Variants using ESM2 and UMAP

Protein sequences can vary due to mutations in their coding DNA sequence, resulting in differences in structure and function. A single protein may exist in multiple variant forms, each potentially leading to distinct phenotypic consequences depending on how the alterations affect its structure, function, or expression. Missense variants are single-nucleotide substitutions in the DNA sequence that cause the replacement of one amino acid with another in the corresponding protein, potentially affecting its structure, stability, and function. The clinical interpretation of missense variants in protein-coding regions remains a challenge in genomic medicine. Recent advances in protein language models and manifold learning offer new opportunities for the unsupervised extraction of biologically relevant information from protein sequences. This study integrates representations derived from Evolutionary Scale Modeling (ESM2), a family of large-scale protein language models, with nonlinear dimensionality reduction via Uniform Manifold Approximation and Projection (UMAP), a technique used to visualize high-dimensional data, to improve the classification of variants of uncertain significance (VUS) in disease-associated proteins. Our results suggest that this approach enhances the separability of benign and pathogenic variants, offering a scalable and interpretable strategy for variant prioritization in precision medicine.