Skip to content

Run an Existing Loci List

Use this tutorial when you already have locus-level files. This is the shortest path into CREDTOOLS.

The Required Files

For each row in your loci list, CREDTOOLS reads files by prefix:

{prefix}.sumstats.gz   # or {prefix}.sumstat
{prefix}.ld.npz        # or {prefix}.ld
{prefix}.ldmap         # or {prefix}.ldmap.gz

Your loci_list.txt should be tab-separated:

locus_id    chr start   end popu    cohort  sample_size prefix
locus_9p21  9   21900000    22100000    EUR UKBB    400000  data/EUR_UKBB_9p21
locus_9p21  9   21900000    22100000    AFR MVP 90000   data/AFR_MVP_9p21

All rows with the same locus_id must use the same chr, start, and end.

Do not include file extensions in prefix

Use data/EUR_UKBB_9p21, not data/EUR_UKBB_9p21.sumstats.gz.

Validate the Shape

Before a large run, inspect a few rows:

head loci_list.txt
ls data/EUR_UKBB_9p21.*

Then ask CREDTOOLS for command help:

credtools pipeline --help

Run the Pipeline

credtools pipeline loci_list.txt results --tool susie

This creates one output directory per locus:

results/
- locus_9p21/
  - pips.txt.gz
  - credible_sets_summary.txt.gz
  - causal_variants.txt.gz
  - run_summary.log
- overall_run_summary.log

Choose the Meta Method

Use --meta-method to control how rows are combined before fine-mapping:

credtools pipeline loci_list.txt results_meta_all \
  --meta-method meta_all \
  --tool susie

Use this when one combined analysis is the goal.

credtools pipeline loci_list.txt results_by_population \
  --meta-method meta_by_population \
  --tool susie

Use this when you want population-level inputs before fine-mapping.

credtools pipeline loci_list.txt results_no_meta \
  --meta-method no_meta \
  --tool multisusie

Use this when a multi-input tool should see the rows separately.

Check the Main Result

Read the PIP table:

gzip -cd results/locus_9p21/pips.txt.gz | head

Read the credible set summary:

gzip -cd results/locus_9p21/credible_sets_summary.txt.gz | head

If no credible set is reported, check QC and the smallest p-values in the locus. No credible set can be a valid result when the evidence is weak.