Svante Pääbo's team uses AlphaGenome as a time machine to reconstruct human gene regulatory evolution

gleech · x · 2026-10-24

A bioRxiv preprint from Svante Pääbo and Henrik Kaessmann's groups uses deep learning (AlphaGenome) to test the 50-year-old hypothesis that human-chimp differences lie mostly in gene regulation. Key findings:

Over 71,000 genomic regions predicted to have gained regulatory activity along the human lineage
In 71% of cases, evolution repurposed existing regulatory elements rather than creating new ones — a single mutation could activate an existing element in another tissue via a new TF binding site
AlphaGenome predicted chromatin accessibility across 305 cell types, validated with cross-species ATAC-seq; it outperformed Enformer and ChromBPnet baselines (AP 0.366)

Original post →

Svante Pääbo's team uses AlphaGenome as a time machine to reconstruct human gene regulatory evolution

More from AGI Musings