Is there any way to output the variant_id or rsID for the SNPs used in S-PrediXcan ？

[Stray357]

Hi, I'm using S-PrediXcan with MASHR model on summary GWAS data following the tutorial at: https://github.com/hakyimlab/MetaXcan/wiki/Tutorial:-GTEx-v8-MASH-models-integration-with-a-Coronary-Artery-Disease-GWAS

Everything goes fine from harmonization to imputation, and I successfully get the output with headings like (NA column omitted):

However, I can only see the number of SNPs used in the model. Is there any way that I can see the exact SNP rsID or variant ID used in the model?

Thanks a lot!

[Heroico]

Hi there!

The base SPrediXcan implementation doesn't output this.

Since prediction model files (.db) are sqlite files, you can identifiy this using an interactive scripting environment like R or Python with the adequate packages. At a conceptual level:

• Load the sqlite file
• query the weights table. For every gene, this has a row for every variant in its model
• load your GWAS file
• perform an inner join by variant id
  • If you used the harmonization script in this repository, variant ids in your GWAS file are compatible
  • If not, I recommend you keep only variants that have the same effect and non-effect allele on both model and GWAS (or at most where the effect and non-effect allele are swapped)

[Stray357]

Thank you very much for the reply.
Just to be sure, may I understand that every SNP in the GWAS with a match in the MASHR model would be included?