Transcriptomic risk scores for attention deficit/hyperactivity disorder

Abstract

Attention deficit/hyperactivity disorder (ADHD) is a highly heritable neurodevelopmental disorder. We performed a transcriptome-wide association study (TWAS) using the latest genome-wide association study (GWAS) meta-analysis, in 38,691 individuals with ADHD and 186,843 controls, and 14 gene-expression reference panels across multiple brain tissues and whole blood. Based on TWAS results, we selected subsets of genes and constructed transcriptomic risk scores (TRSs) for the disorder in peripheral blood mononuclear cells of individuals with ADHD and controls. We found evidence of association between ADHD and TRSs constructed using expression profiles from multiple brain areas, with individuals with ADHD carrying a higher burden of TRSs than controls. TRSs were uncorrelated with the polygenic risk score (PRS) for ADHD and, in combination with PRS, improved significantly the proportion of variance explained over the PRS-only model. These results support the complementary predictive potential of genetic and transcriptomic profiles in blood and underscore the potential utility of gene expression for risk prediction and deeper insight in molecular mechanisms underlying ADHD.

Fig. 1. Observed and predicted (multi-tissue TWAS) differential expression for ADHD.

Z-scores are plotted for the 56 genes significantly associated with ADHD in at least one of the studied tissues. Significantly associated genes are outlined in black (TWAS results with P < 8e-06, and observed expression in blood FDR < 0.05). The dots are color-coded based on the Z-scores of the genes, with white indicating a Z-score of 0, blue indicating a negative Z-score, and red indicating a positive Z-score. In the y axis novel genes that were not previously identified in the original ADHD GWAS or TWAS on the dorsolateral prefrontal cortex by Demontis et al., 2022 are highlighted in bold. * Genes previously reported in the GWAS and TWAS on ADHD by Demontis et al., 2022. ** Genes previously reported in the TWAS on ADHD by Demontis et al., 2022.

Fig. 2. Quintile plot of odds ratios for PRS and TRS.

Odds ratios (OR) with 95% confidence intervals are shown for PRS and TRSs that overcome multiple comparison corrections and sensitivity analysis using the first quintile as baseline.

Fig. 3. Proportion of variance explained by TRSs significantly associated with ADHD and by PRS.

Pseudo-R² (in the y-axis) is presented for each TRS, TRS restricted to colocalized genes (TRS_col), the PRS and the model that combines both scores (TRS + PRS or TRS_col + PRS). Values in brackets indicate the best P-value threshold for a given tissue (P). Likelihood ratio test P-values for TRS and PRS vs. PRS-only model comparisons are given above the bars. Further statistical details can be found Supplementary Table S9.