We then sought to extend the power for gene discovery, as well as supporting signals uncovered by the TWAS, through integration of models of genetically regulated protein expression (PWAS) and alternative splicing (spliceWAS, online Supplementary Tables S2, S3). In other words, these analyses considered the association of genetically predicted protein expression and alternative splicing with AN rather than mRNA expression. There were 2300 protein expression SNP weight sets in total available to test across the DLPFC and blood cohorts in which the imputed expression models were trained. Two proteins survived Bonferroni correction in the PWAS, and we uncovered four signals that were significant after applying a less stringent FDR based cut-off (FDR < 0.05). The strongest protein association, MHC class I chain-related gene B (MICB) (ZPWAS = 4.53, p = 5.75 × 10−6, plasma) was correlated with increased odds risk of AN, however, due to the extensive LD of the MHC region it is difficult to apply approaches such as PWAS, and thus, this signal should be treated cautiously. The densely associated region on chromosome 3 harboured the next most
