Fig. 3: Candidate markers predicting general IQ.
From: Epigenetic variance in dopamine D2 receptor: a marker of IQ malleability?
For display purpose, we grouped individuals into septiles of the candidate markers and plotted the mean phenotypic value (here general IQ) for each quantile on the y-axis52. Error bars indicate standard error of the mean. a General IQ can be predicted using polygenic score from Sniekers et al.4 at a p-threshold of 0.01 comprising 5636 SNPs explaining 3.2% of variance (df = 1376; p = 7.3 × 10−8; correcting for age, gender, study site, principal components from imputation, and genetic strata). b Here we display association with the marker with the lowest p-value (methylation count in dopamine D2-receptor gene, DRD2 cg26132809) among our candidate markers. We grouped individuals into septiles of their methylation level (higher septile rank indicating higher probability of methylation) and plotted those septiles against mean general IQ score on the y-axis. General IQ is negatively correlated with candidate marker for dopamine neurotransmission in our regression model (2.7% of variance explained, df = 803, p = 3.18 × 10−4 correcting for age, gender, study site, wave information, and variability in cell type) indicating that higher methylation count, which is considered as downregulation of transcription of DRD2 receptor, is related to lower IQ scores. c Gray matter density in bilateral striatum was used to group individuals into septiles. We plotted gray matter density against general IQ and found 0.71% variance explained (df = 1399, p = 1.7 × 10−3), correcting for age, gender, site, and total brain volume. d Here we plot general IQ by reward anticipation signal (BOLD-signal) in region of interest (ROI). We grouped individuals into septiles of beta parameter estimates (BOLD-signal) and plotted mean general IQ for each quantile on the y-axis for display purposes. General IQ is positively correlated with functional activation of the ventral striatum (1.4% of variance explained, df = 1463, p = 4.11 × 10−6; correcting for gender, age, and study site). e Regression model illustrating neurobiological correlates of general IQ in an overlapping sample of n = 755. A multiple linear regression model with general IQ (gIQ) as outcome variable was estimated with the residuals of the following predictors: polygenic score (from Sniekers et al.), methylation in DRD2 gene, gray matter in striatum, and functional activation during reward anticipation. The whole model was significant with an adjusted R2 = 0.04 (df = 750, p = 3.3 × 10−7). On the edges, we display the standardized parameter estimates for each predictor (beta) describing how many standard deviations the dependent variable (gIQ) will change, per standard deviation increase in the predictor variable. With respect to the different predictors, we could replicate previous findings that the established polygenic score (including 5636 SNPs significant at a p-threshold of 0.01) shows an association with general IQ (beta = 0.13, p = 2.8 × 10−4). We find variance in methylation count in our candidate CG site (DRD2 cg26132809) that is negatively associated with general IQ (beta = −0.10, p = 6.2 × 10−3), indicating that higher methylation (lower gene activity) being associated with lower gIQ. In this subsample gray matter density in striatum was not associated with gIQ (beta = 0.02, p = 0.5). BOLD-signal change during reward anticipation significantly predicts cognitive capacity (beta = 0.14, p = 9.4 × 10−05)
