GPN-Star: evolution-aware genome model hits SOTA, published in Nature

EricTopol · x · 2026-09-09

UC Berkeley researchers published GPN-Star in Nature: a phylogeny-aware genomic language model that explicitly uses whole-genome alignments and species trees to model evolution. Trained across vertebrate, mammal and primate timescales, it achieves state-of-the-art variant effect prediction in both coding and non-coding regions — outperforming larger NLP-style genomic models and classical evolutionary methods — moving toward predicting the molecular impact of all 9 billion human genome variants.

Related event: Berkeley's GPN-Star Genomic Model Published in Nature(2 posts)→

Original post →

More from Research

Research channel →