calculated with those same HapMap3 SNPs. The TWAS using whole blood weights utilised the same LD reference strategy, with 369 GTEx v7 participants in these models. Frontal or cerebral cortex tissue from the larger PsychENCODE cohort was also utilised in terms of SNP weights, with a sample size of 1695. As described in the original PsychENCODE publication, gene-wise GReX were estimated for all imputed SNPs, not just the HapMap3 panel, and thus, we utilised the full suite of the phase 3 1000 genomes European subset as the LD reference. There were two tissues for which SNP weights related to protein expression were available – the dorsolateral prefrontal cortex (DLPFC, N = 376) and plasma (N = 7213) (Wingo et al., 2021; Zhang et al., 2021). It should be noted that the plasma weights were derived from the European subset of the study cohort and the authors only used the elastic net method to derive GReX. Analogous to the difference between the GTEx v7 and PsychENCODE studies above, the DLPFC SNP weights were estimated using the HapMap3 panel, and thus, we only used those SNPs as an LD reference. The plasma PWAS utilised the full reference panel. Finally, the splicing related
