SuSiE fine-mapping

[dariushghasemi]

The scripts and steps followed to run susie using in-sample LD on the meta-analysis GWAS results:

• Computing LD and converting to matrix format for running SUSIE (Fri, 23:37, 21-Jun-24).
• Susie was successfully run on the example protein and finally three causal variants were identified (Sunday, 16:46, 24-Jun-24).

[dariushghasemi]

If there is a misalignment of alleles in the PLINK2 LD file and GWAS results, then the credible set variants will be infinity,

To solve this, we computed the LD matrix with Pearson correlation in R using the dosage exported from PGEN file. The PGEN-based LD looks aligned with GWAS summary file (checked the diagnostic plot of eigenvalues susieR::kriging_rss()). As a result, the susie_rss() model identified one credible set pertaining 6 variants at this locus: seq.12593.33/19_54219624_54427869.

Next, I test it for the other locus with 34 signals!

Thu, 19:18, 11-Jul-24.
Dariush