Speaker
Description
SARS-CoV-2 is characterized by sequential emergence of highly divergent and highly transmissible variants, denoted as variants of concern (VOCs). In parallel, it has been observed that a subset of patients, who suffer from immunosuppression, develop chronic infections. In some case highly divergent variants emerge in these patients with mutations similar to those in VOCs. The similarity between the mutational patterns in VOCs and chronic infections has led to the hypothesis that in some rare cases, chronic infections are the reservoir from which VOCs emerge. However, it remains unclear what are the determinants that allow variants in some chronic infections to become highly transmissible.
Leveraging millions of SARS-CoV-2 sequences, we identified 271 'chronic-like' clades, suspected to represent longitudinal sampling from chronically infected patients. This study aims to test for positive epistasis evidence between mutation pairs in chronic infections. Analyzing mutational pathways, we assessed whether mutation pairs occur more frequently than expected, focusing on mutations with inferred negative fitness and examining the common emergence order.
Preliminary analysis identified at least 29 mutation pairs displaying potential positive epistasis. Notably, the pair S:G446V+S:G476D suggests a combined effect enabling balanced ACE2 binding and resistance to antibody neutralization. Additionally, E:T30I exhibits potential epistasis with various mutations, prompting investigation into its potential ability to infect alternative tissue types.
Ultimately, my plan involves employing a Large Language Model to infer a larger number of chronic-like infections and dependencies between mutations indicative of epistasis. This model-driven approach enables a comprehensive analysis, unveiling hidden patterns and relationships within the extensive SARS-CoV-2 sequence dataset.
