Scaling genetic discovery through automated post-GWAS interpretation

VariantAgent, part of Post-GWAS Intelligence (PGI), achieved mean per-study recovery of 82.1% across 34 published GWAS, with biggest gains when requiring both correct answers and valid analytical traces.

• Why it matters: Interpreting GWAS loci is selective, heterogeneous, and hard to accumulate across studies, limiting how association discoveries translate into usable biological evidence.
• What they did: PGI couples context-dependent post-GWAS analysis to standardized, variant-centred evidence units, using a multi-agent engine (VariantAgent) to generate trait-level interpretations linked to analytical basis and evidential status.
• The result: Across 1,041 GWAS, PGI produced 530,949 evidence units and reports, organizing them into a cross-trait catalogue that resolved apparent overlap into distinct signals and trait-specific biological hypotheses.