Smart Health
SARS-CoV-2 Variant Analysis
This work develops a machine learning pipeline for viral sequence classification with a particular focus on accurate SARS-CoV-2 variant identification at scale.
Back to Research
Project Summary
Sequence classification pipelines using transformer embeddings including ESM-2 and ProtBERT.
Current approaches can lack the precision needed to distinguish closely related SARS-CoV-2 variants within massive genomic datasets, which limits effective public health response and epidemiological surveillance. Create a comprehensive machine learning tool for classifying viral sequences, with specific emphasis on SARS-CoV-2 variant analysis. The modeling pipeline uses protein language models such as ESM-2 and ProtBERT to derive strong sequence representations. The project is designed for large genomic datasets where subtle differences between variants matter operationally. A major aim is to improve the precision gap in identifying closely related variants.