Genetic imputation models for 2,594 plasma proteins — built from 54,219 UK Biobank participants — were projected onto over 640,000 individuals across the UK Biobank and All of Us Research Program, spanning six genetic ancestries. The resulting proteome-wide association study uncovered approximately 9,000 protein-disease associations across 89 cardiovascular conditions. Critically, in vivo CRISPR disruption of liver-derived ITIH4 in a preclinical animal model reduced plasma cholesterol and pro-atherogenic lipid species, providing experimental causal evidence for a computationally nominated target.
This work represents a methodological step-change rather than an incremental finding. Most proteomics studies are constrained to tens of thousands of participants with measured proteins; genetic imputation scales that reach to hundreds of thousands while preserving ancestry diversity — a persistent blind spot in cardiovascular genomics. The consistent cross-ancestry and cross-biobank replication of the majority of associations meaningfully strengthens confidence in the proteomic signatures identified. The ITIH4 finding is particularly compelling: moving from statistical association to CRISPR functional validation in one study is rare and narrows the target-to-drug pipeline considerably. Limitations include reliance on imputed rather than directly measured proteomes, which introduces model error, and the preclinical CRISPR work is animal-based with uncertain human translation. Broad population effects of ITIH4 disruption remain unknown. As a preprint posted to medRxiv and not yet peer-reviewed, findings — especially mechanistic claims — should be treated as provisional until independent scrutiny is complete.
