Summary of Pipeline Updates

1. Utility Function

   • Added a new function z_to_pvalue() in misc.R, which is now used twice in univariate_rss_diagnostics.R.

2. Diagnostics Functions

   • Included all functions used for diagnostics table in univariate_rss_diagnostics.R for modularity and reusability.

3. Pipeline Modification

   • Modified rss_analysis_pipeline() to include a new parameter:
     diagnostics (default: FALSE).

   • When diagnostics = TRUE, the pipeline performs additional steps:

     Step 1: Generate Diagnostics Table

     • Extract and process information from the first finemapping analysis, which should satisfy in this automated QC:
       • finemapping_method == "susie_rss"
       • qc_method == "rss_qc"
       • impute == TRUE
     • The output is a diagnostics table summarizing study-block level results.

     Step 2: Trigger Additional Analyses Based on Diagnostics

     • For each study_block:

       • If CS > 1:

         • Check for any spurious credible sets, defined as:
           p_value > 1e-4 or cs_corr_max > 0.5
         • If spurious CS exists:
           • Run BCR (impute, rss_qc)
           • Run SER (non_impute, noqc)
         • If not spurious:
           • Keep the original BVSR result (from first analysis)

       • Else (CS == 1 or no CS found):

         • Run SER (non_impute, noqc)
         • Note: I recommend not replacing the first finemapping result with the new SER output, as some BVSR results may be more informative.

   • I modified the naming of method layer to distinguish the orginal and additional results. I didn't store the sumstats after impute and qc for additional analyses.

4. Documentation

   • Updated the documentation to reflect these changes.

I tested using chr11_84267999_86714492 data in AD_GWAS summary statistics, it worked.