Removing genetic effects on plasma proteins enhances their utility as disease biomarkers

Abstract

Abstract Plasma protein levels are influenced by both genetic and non-genetic factors and can serve as early disease biomarkers. When a protein is correlated with, but not causally linked to, disease, its genetic determinants can add unwanted variability to protein–disease associations. In such cases, removing the genetic component may improve their predictive performances. Here, we tested this hypothesis by genetically adjusting 94 highly heritable proteins spanning diverse biological pathways and evaluating their associations with the onset of 37 diseases in 39,871 UK Biobank participants. Genetically adjusted proteins showed stronger associations in 88% of 1,312 significant protein–disease pairs, equivalent to a 30% median reduction in required sample size for comparable power. Of 96 protein–disease pairs with significant differences, all but one showed larger effects for adjusted proteins. Most proteins also showed consistently stronger associations with environmental and lifestyle factors once genetic effects were removed. Finally, we constructed multi-protein scores from genetically adjusted proteins and demonstrated that they significantly improve prediction for 7 diseases compared to unadjusted proteins. These findings demonstrate that removing genetic effects from plasma proteins is an effective strategy to increase power for biomarker discovery and clinical trial design, consistent with the largely non-causal role of most plasma proteins in disease risk.