Summary level statistics from the Genomic SEM analysis were used as input to MetaXcan. Prior to input, summary statistics were harmonized according to the best practices guide outlined on the MetaXcan wiki. As part of this process, the gwas_parsing.py utility (https://github.com/hakyimlab/summary-gwas-imputation) was used to lift summary statistics over to the human genome build version 38 and provide harmonized variant identifiers compatible with those used by GTEx v8. To increase the number of overlapping markers between our summary statistics and the fine-mapped pre-built MASHR models, we imputed missing summary associations as suggested by the best practices workflow. Imputation was performed separately for each chromosome using the gwas_summary_imputation.py utility (https://github.com/hakyimlab/summary-gwas-imputation) and the pre-computed parquet genotype, genotype metadata files, and European LD block files available through the MetaXcan zenodo repository76. Imputed summary statistics were finally re-combined using the gwas_summary_imputation_postprocess.py utility.
