The proportion of the phenotypic variance that could be explained by the aggregated effect of all included SNPs (SNP-based heritability, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${h}_{{\rm{SNP}}}^{2}$$\end{document}hSNP2) was estimated using LDSC31. The analysis was performed using precomputed linkage disequilibrium scores from samples restricted to European ancestry in the 1000 Genomes Project95, filtered for SNPs included in the HapMap 3 reference panel101. SNP heritability was estimated based on the slope of the LDSC, with heritability on the liability scale calculated assuming a 1% population prevalence of OCD1. To omit a downward bias in our estimates of liability-scale heritability, following Grotzinger et al.102, we accounted for varying levels of ascertainment across cohorts in our meta-analysis by summing the effective sample sizes across the contributing cohorts and using that as the input sample size for LDSC. For conversion to the liability scale (1%), the sample prevalence was then specified as 0.5. The SNP heritability was calculated for the whole OCD sample as well as for ascertainment-specific subgroups.
