VENAS2: Bidirectional Mutation Modeling for Genomic Surveillance Reveals Reversions and Hidden Transmission Links in Pathogens

Abstract

Accurately resolving pathogen evolution is essential for genomic surveillance, yet existing approaches often assume unidirectional mutation accumulation and fail to capture reversions that obscure lineage classification and transmission dynamics. Here we present VENAS2, a scalable framework that integrates genomic, temporal and geographic information to construct spatiotemporal evolutionary networks while explicitly modeling bidirectional mutational processes. Applied to 7.25 million SARS-CoV-2 genomes from GISAID between December 2019 and June 2024, VENAS2 reconstructed the evolutionary history of SARS-CoV-2 and precisely inferred the temporal dynamics of major variants of concern. Further application to the Viridian dataset revealed 4,654 distinct genomic sites with at least one potential reversion event, providing a systematic view of potential reversion signals during viral evolution. Extending to Mycobacterium tuberculosis, VENAS2 delineated major lineages and identified over threefold additional transmission clusters beyond conventional genomic distance-based methods, uncovering hidden links enriched for drug-resistant strains. These results demonstrate VENAS2’s capacity to uncover cryptic evolutionary processes across viruses and bacteria, providing a generalizable platform that bridges genomic surveillance with actionable public health response.