A genomic foundation model for interpreting human genetic diversity

Genomic foundation model pretrained on haplotype sequences reconstructed from WGS, without clinical or functional labels

hEVO is a genomic foundation model trained on 4.784 billion distinct haplotype sequences reconstructed from WGS data for 63,809 individuals selected from 550,237 samples across eight cohorts. Pretraining integrates shared sequence features and naturally occurring allelic differences without clinical or functional labels. Against representatives of the Evo 2, NT, GPN and other genomic model families, hEVO ranked first in 18 of 19 primary comparisons spanning functional annotation, variant interpretation and human evolution. Without task-specific training, its allele scores achieved the highest performance across six variant benchmarks, including clinical pathogenicity, and a retrospective ClinVar reclassification showed improved identification of pathogenic classifications later downgraded.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

Preprint, not peer reviewed. For the laboratory, the most telling result is the last one: a score able to flag pathogenic classifications that will later be downgraded becomes a check on our own past interpretations. The abstract, however, does not quantify this gain or state whether the model is available; the gap between benchmark performance and real benefit for ACMG classification remains to be measured.

Analysis by Dr Thibaut Benquey