AI‑Phenotypes Reveal Selection Signals

Using deep‑learning to derive standardized, high‑quality phenotypes (e.g., retinal pigmentation from fundus photos) removes a key bottleneck in large‑scale GWAS and lets researchers test polygenic selection with phenotypes that are consistent across cohorts. Coupled with explicit demographic covariance models (Qx), AI‑phenotyping can make within‑region selection tests more robust to ancestry confounding. — If generalized, AI‑derived phenotypes plus strict provenance and structure controls change how we detect recent selection, that will affect public debates about genetic differences, the clinical use of PGS, and standards for reproducible human‑genetics claims.