Williams · medRxiv : the preprint server for health sciences 2026 · multi-ancestry multivariate genome-wide association study · n=>4,000,000

Genomic insights into substance use and disinhibitory disorders.

Level 3 - non-randomized controlled study

Large-scale observational multivariate genome-wide association study across population cohorts (graded by design analogy).

PubMed 41728329 · doi:10.64898/2026.02.09.26344198 · record verified 2026-08-26

What was done

Researchers conducted a multi-ancestry, multivariate genome-wide association study (GWAS) involving more than 4 million individuals to investigate the shared genetic architecture underlying externalizing spectrum traits (including ADHD, conduct disorder, and substance use disorders). They performed fine-mapping and gene prioritization, evaluated replication of putative causal variants in the All of Us Research Program, conducted bioinformatic and drug-repurposing analyses, and assessed the performance of a genome-wide polygenic index across different ancestral populations.

What was found

The analysis identified 1,294 genomic regions and prioritized 961 effector genes associated with the broad externalizing factor. Functional and bioinformatic analyses implicated GABAergic and glutamatergic neurons, GABA_A receptors, dopaminergic signaling, neurosteroid pathways, and excitatory-inhibitory balance. The derived polygenic index explained approximately 12% of the phenotypic variance in externalizing traits in independent cohorts of European-like ancestry, compared to approximately 3% in individuals of African-like ancestry.

Why it matters

This study provides the largest genomic map to date of shared liability across disinhibitory and substance use disorders, pinpointing specific synaptic and neurodevelopmental pathways for potential therapeutic exploration. It also highlights a substantial ancestry-based disparity in the clinical and research utility of current polygenic scores.

Limits

The abstract provides summary metrics but omits specific individual effect sizes and confidence intervals. A major disparity in polygenic prediction remains, with predictive utility dropping fourfold in African-ancestry samples compared to European-ancestry samples. Biological pathway and drug-repurposing insights are derived from statistical and bioinformatic modeling rather than direct experimental or functional validation. Additionally, collapsing clinically diverse disinhibitory conditions into a shared externalizing factor may obscure phenotype-specific etiologies.

Cited by