CoVFit

A protein language model for accurately predicting the fitness of SARS-CoV-2 variants

Summary

Continually emerging SARS-CoV-2 variants lead to repeated epidemic surges through escalated spreading potential (i.e., fitness). This repository hosts the software and code of CoVFit, a protein language model able to predict the fitness of newly emerged variants based solely on their spike protein sequences.

Contents

CoVFit-CLI

• The CoVFit models can be used as a stand-alone command line tool to predict the viral fitness of SARS-CoV-2 spike protein sequences in fasta format. Download link and instructions are available here.

CoVFit Training

• Training
• Figures
• Dataset

Domain Adaptation

• Code used for performing domain adaptation to produce the ESM2_{coronaviridae} model and a download link for the model itself are available here.

ML Comparison Models

• Description
• Figures

Citation

A Protein Language Model for Exploring Viral Fitness Landscapes. Jumpei Ito, Adam Strange, Wei Liu, Gustav Joas, Spyros Lytras, The Genotype to Phenotype Japan (G2P-Japan) Consortium, Kei Sato. 2024. bioRxiv https://doi.org/10.1101/2024.03.15.584819

Contact

jampei@g.ecc.u-tokyo.ac.jp (Jumpei Ito)