Genetic studies on psychological traits often adopt phenotypic composite scores, such as a sum-score or binary case–control status, which summarize information contained in multiple items or symptoms. In psychological research, composite scores have proven useful, e.g., in directing therapeutic intervention, and in predicting future school/job performance. However, the items or symptoms collectively operationalizing one trait can be very diverse in nature. For instance, in personality inventories for neuroticism, items vary from ‘feeling miserable’ and ‘experiencing mood swings’ to ‘feeling guilty’ and ‘worry too long after an embarrassing experience’. Similarly, diagnostic symptoms for major depressive disorder (MDD) vary from ‘increase in appetite’, ‘irritable mood’, and ‘fatigue’ to ‘insomnia’, ‘excessive guilt’, and ‘suicidal ideation’6. Previous studies7–9 have shown that items or symptoms underlying the same composite score can indeed differ considerably with respect to e.g. their relations to external risk factors, their impact on impairment, and their underlying biology. In the context of gene-finding studies, power to detect associated variants is potentially lost when the summed items or symptoms are biologically heterogeneous. Specifically, the use of composite scores in gene-finding studies directs
