Dissecting polygenic risk for neurodevelopmental disorders by educational attainment and exploring their relationship with school performance

Neurodevelopmental disorders like attention-deficit/hyperactivity disorder (ADHD) or autism spectrum disorders (ASD) are strongly associated with school performance and educational attainment (EA). However, genetic correlations between ADHD or ASD and EA show opposite patterns. We aim to test the association between genome-wide polygenic scores (PGS) for ADHD or ASD dissected by their relationship with EA in school performance and assess whether these associations are mediated by ADHD or ASD symptoms among 4,278 schoolchildren from the INSchool cohort. School performance was assessed as school grades in three different domains: mathematics, English and language. PGS for ADHD and ASD were constructed using PRS-CS and summary statistics of GWAS meta-analyses on EA, ADHD and ASD. For polygenic dissection we used subsets of SNPs based on their contribution to EA: (i) the overall set of SNPs, (ii) variants not associated with EA; (iii) variants associated with EA; (iv) variants associated with EA and showing consistent direction of effects in EA and ADHD or ASD (PGSconcordant) and (v) variants associated with EA showing opposite direction of effects in EA and ADHD or ASD (PGSdiscordant). Associations between PGS and school performance were assessed using ordinal regression models with age, sex, socioeconomical status and 20 PCs as covariates and school as random effects. PGS for ADHD, but not for ASD, were associated with poor performance in mathematics, English and language. PGS_ADHDdiscordant (including variants associated with ADHD showing opposite direction of effects in EA and ADHD) were associated with worse school performance while PGS_ADHDconcordant were not. PGS comparisons across ranked quintiles showed the expected trend, with children at the highest quintile of PGS_ADHDdiscordant doubling the odds for performing worse at school compared to children in first quintile. When we subset the PGS for ASD based on its role in EA we found opposite direction of association, with PGS_ASDdiscordant and PGS_ASDconcordant associated with poor and better school performance in all three domains, respectively. PGS comparisons across ranked quintiles showed that children at the highest quintile for PGS_ASDdiscordant had around 1.5 times more odds for poor performance than children in the first quintile, while children at higher quintiles for PGS_ASDconcordant showed better performance than children in the first quintile in all three domains. These associations between PGS_ADHD and PGS_ASD and school performance were mediated, in part, by ADHD or ASD symptoms, respectively. The association between PGS for neurodevelopmental disorders dissected by their relationship with EA also provided distinct patterns of behavioral and emotional problems between the two groups. Overall, behavioral and emotional problems were positively correlated with PRS constructed on variants that increase the risk for ADHD or ASD and discordant effects for EA. Our results reveal that, despite common genetic architecture, the genetic liability for ADHD and ASD shows differences in their polygenic association with school performance when considering its relationship with EA: genetic variation with discordant effects in ADHD or ASD and EA contributes to poor school performance, while genetic variation with concordant effects in ASD and EA are associated with better school performance.