Using Gene Set Enrichment Analysis to Create Polygenic Risk Scores for Childhood Externalizing Behavior

GWAS-based polygenic risk scores (PRS) for childhood externalizing behavior have predicted aggression, impulsiveness, and conscientiousness (Salvatore et al., 2015). Selection of SNPs for PRS is typically statistically-driven,... [ view full abstract ]

GWAS-based polygenic risk scores (PRS) for childhood externalizing behavior have predicted aggression, impulsiveness, and conscientiousness (Salvatore et al., 2015). Selection of SNPs for PRS is typically statistically-driven, irrelevant of biological mechanisms. We used gene set enrichment analyses (GSEA) to inform selection of SNPs from a discovery GWAS to create PRS from biologically-related genes, and tested associations with childhood externalizing behavior. Results from a Meta-GWAS of childhood externalizing behavior (Pappa et al., 2016) were mapped to gene sets containing 10-200 genes using iGSEA4GWASv2 (Zhang et al., 2015). Using SNPs at gene-set significance levels of FDR p<.01, p<.05, p<.10, and p<.20 (136 to 252 SNPs), PRS were created. PRS were also created for SNPs from the original GWAS at the same significance levels (1570 to 29,433 SNPs).

Participants were from a longitudinal randomized prevention trial of high-risk children. Children in the current subsample were 2 to 5 years old (n = 391), 48% female and 55% Caucasian, 29% African American, 13% Hispanic, 3% Other. Parents reported on the externalizing subscale of Achenbach’s CBCL in each year. A latent measure of aggression was created across ages 2-5 (RMSEA = .04, CFI = .99). Analyses controlled for child age, ethnicity, gender, and intervention condition.

PRS created from the original GWAS were individually tested at each significance level and none were associated with childhood externalizing behavior (ps>.16, R²=.001 to .009). GSEA-scores at p<.05, p<.10, and p<.20 significance levels were significantly associated with childhood externalizing behavior (β=.11*, β=.13*, β=.12*, respectively), each explaining 1-2% of variation in behavior. Additional analyses will be presented on prediction of trajectories of externalizing behavior.

Although these preliminary results require replication, using GSEA-informed polygenic risk explained more variation in child externalizing behavior than statistically-driven polygenic risk. GSEA is a promising method for advancing understanding of genetic association with quantitative traits.