effects mediated through rare or structural variants. Future larger-scale AN GWAS planned will help to consolidate the strength and replicability of these findings. Previous work in power analysis for TWAS approaches have suggested that both expression and trait heritability/sample size influence discovery power. The liability scale SNP heritability of AN even at current sample sizes is larger than several other psychiatric disorders, supporting that post-GWAS analyses can be deployed despite the above limitations. Moreover, expression-based approaches would be greatly improved by access to cell-type specific genetic models of expression. In terms of bulk-tissue panels, the PsychENCODE cortical dataset is well-powered and captures genetic effects on numerous genes, however, this is not the case for other brain regions and tissues in GTEx where sample size is much smaller. Another limitation of the eQTL association data used in this analysis is that it captures steady-state expression levels which cannot directly distinguish between decay rates of mRNA and transcriptional variance (Pai et al., 2012). Furthermore, these analyses were performed in exclusively European ancestries, and as more diverse non-European samples and trans-ancestry GWAS become available, this is likely to improve TWAS and finemapping studies (Aguet et al., 2017; Pai et al., 2012; Veturi &
