Multi-rater and longitudinal GWAS of Attention and Aggression during development and the genetic correlations with adult psychiatric outcomes

We present the first results of a meta-analysis on childhood behavioral problems involving a collaboration between 17 international cohorts, form Europe (Finland, Germany, Netherlands, Spain, Sweden, UK), Australia,... [ view full abstract ]

We present the first results of a meta-analysis on childhood behavioral problems involving a collaboration between 17 international cohorts, form Europe (Finland, Germany, Netherlands, Spain, Sweden, UK), Australia, New-Zealand and the USA. The cohorts collaborate within the ACTION and EAGLE consortia, and are characterized by rich phenotyping in children, includingrepeated measures of aggressive behavior (AGG) and attention problems (ATT).

We analyzed longitudinal data of children and adolescents aged 3 to 18 years, assessed by multiple raters (mother, father, teacher, and self) and multiple instruments. First, a series of univariate GWAS’s was performed for every available combination of age and rater within each cohort, resulting in 3 to 36 analyses per cohort. Genotypes were imputed using the 1000 genomes Phase III reference set and mapped to the human reference genome GRCh37. Second, results were pooled into age by rater combinations (e.g. mother-rated aged 3-5, teacher-rated 8-11, etc.) that resulted in an excess of 10.000 observations and then meta-analyzed. We performed inverse variance weighted meta-analysis in R using custom code, correcting for the fact that repeatedly measured subjects were included in the analyses. We then used LD score regression to estimate the genetic correlation between the various rater by age combinations, across AGG and ATT. Furthermore, the genetic correlation between results of the meta-analyses and adult psychiatric outcome was estimated in order to determine the relative information found in ratings made by the various raters at different ages. Finally, we performed a meta-regression analysis across all GWAS’s. In total, over 200 GWAS’s were included in the meta-regression analysis, where the total number of observations equaled ~526.000 (the range of observations for individual age by rater by instrument GWAS was 309 to 10,812).