A deep mutational scanning-informed protein language model predicts SARS-CoV-2 evolution dynamics with spatiotemporal resolution - PubMed
2 hours ago
- #SARS-CoV-2 evolution
- #variant prediction
- #deep mutational scanning
- DeepCoV is a deep-learning framework that predicts emerging SARS-CoV-2 variants with spatiotemporal resolution by integrating DMS data, evolutionary sequences, and epidemiological surveillance.
- It forecasts variant dominance a month in advance with 90% reduced false discovery rate, capturing temporal and geographic spread dynamics and reconstructing regional prevalence trajectories.
- The model identifies mutational hotspots in Omicron-derived backbones, revealing convergent evolution trends to provide actionable insights for timely detection of immune-evasive variants.