We used both Transcriptome-wide Association Study (TWAS)49 and Summary-based Mendelian Randomisation (SMR)51 methods to infer differential gene expression associated with MDD based on the meta-analyzed summary statistics. Both methods test whether genetic variants associated with MDD are also associated with differential expression of nearby genes. TWAS and SMR have different limitations and are therefore complementary. TWAS considers the effect of multiple variants on gene expression and the GWAS phenotype, thereby increasing statistical power to detect associations, whereas SMR only considers the effect of each variant individually. However, TWAS requires multi-variant models predicting gene expression to have been generated in the genotype-expression dataset, which are not available in some cases. In contrast, SMR requires only expression quantitative trait (eQTL) summary statistics, enabling it to use a wider range of genotype-expression datasets, such as eQTL meta-analysis results from eQTLGen77 and MetaBrain78 consortia. For TWAS and SMR analysis, the European subset of the 1000 Genomes Project, Phase 3 was used as an LD reference.
