Argonne rewrites genomics code for GPUs: 2,068 traits across 635,969 veterans

Argonne National Laboratory rewrote the SAIGE genomics package for GPUs, cutting an analysis that took roughly 481,000 CPU-core hours. Using Million Veteran Program data, the team studied 2,068 traits across 635,969 veterans and flagged 13,672 disease-linked genomic regions, about 1,600 of which appeared only thanks to non-European ancestry in the cohort.

• SAIGE-GPU replaced a CPU run that took about 481,000 core hours
• Analysis covered 2,068 traits across 635,969 veterans
• 13,672 genomic regions linked to conditions were flagged
• About 1,600 associations came only from non-European groups