To investigate the transcriptome-wide associations between predicted gene expression and OA, we employed the MetaXcan v0.6.642 method. Briefly, MetaXcan uses association summary statistics to predict associations between gene expression and a phenotype of interest association. Gene expression models were predicted from tissue-specific eQTL datasets. To increase the performance of our prediction models, we used the MASHR-M73 models built on fine-mapped variables from DAP-G74. The specific models we used were pre-computed MetaXcan models available through PredictDB (http://predictdb.org/) for 12 brain regions (Supplementary Table 14) that were generated using the GTEx75 version 8 datasets.
