Imputation of fluid intelligence scores reduces ascertainment bias and increases power for analyses of common and rare variants

Abstract Studying the genetics of intelligence provides insight into the neurobiology of cognitive function and rare neurodevelopmental conditions. Here we integrated fluid intelligence (FI) tests across different UK Biobank measures and imputed FI for unmeasured individuals, increasing the sample size from ~270,000 to >455,000. We show through extensive validation that the imputed phenotype genetically resembles measured FI, although some noncognitive signal likely remains. Combining measured and imputed FI increases the number of independent significant common SNP associations from 390 to 550 and reduces the ascertainment bias introduced by FI test participants being, on average, more educated. Rare variant analyses identify 26 significantly associated genes (false discovery rate < 1%), including 8 without strong prior evidence of involvement in intelligence or neurodevelopmental conditions, which replicate in aggregate in external cohorts. Our results demonstrate that imputed intelligence can boost power for genetic discovery and support its use to help discover new neurodevelopmental condition-associated genes.