What question did this study set out to answer?

The aim is to assess the effectiveness of protein language model embeddings for assigning viral clades.

April 23, 2026Journal of Computational Biology0 citations

Evaluating Protein Language Model Embeddings for Viral Clade Assignment

BSBrendonas Stakauskas VMVirginijus Marcinkevičius

Key Points

The aim is to assess the effectiveness of protein language model embeddings for assigning viral clades.
Evaluated embeddings from ProtT5, ProtBert, CARP, and ESM-2 on influenza A/H3N2 sequences.
Applied dimensionality reduction techniques like t-SNE, UMAP, PCA, and MDS.
Used clustering methods, particularly HDBSCAN, to compare against baseline approaches.
t-SNE with PLM embeddings successfully recovered clade structures.
ProtBert showed the most stable performance across evaluations.
Larger ESM-2 models sometimes achieved lower variation scores but exhibited greater variability.

Abstract

Protein language models (PLMs) provide powerful sequence representations, yet their effectiveness for unsupervised viral clade assignment remains uncertain. In this study, we evaluated embeddings from ProtT5, ProtBert, CARP, and several ESM-2 variants on influenza A/H3N2 hemagglutinin sequences. Using dimensionality reduction (t-SNE, UMAP, PCA, MDS) and clustering with HDBSCAN, we compared PLM embeddings against baseline Hamming distance approaches. Our results show that t-SNE combined with PLM embeddings can recover clade structure, with ProtBert yielding the most stable performance and larger ESM-2 models occasionally achieving lower normalized variation of information scores but with greater variability. These findings suggest that while PLM embeddings capture clade-relevant signals, they also suffer from instability and the loss of site- or nucleotide-specific detail. Future improvements in pooling strategies may enhance their utility for viral surveillance.

问 AI

Bookmark

View Full Paper