Choosing a Fine-Mapping Tool¶

Start simple. Use SuSiE for the first pass, then compare other tools for loci that matter.

Quick Choice¶

Situation	Good first tool
standard single-input locus	`susie`
no LD matrix available in Python	`abf`
many cohorts or populations kept separate	`multisusie`
multi-ancestry comparison with separate inputs	`susiex`
sensitivity analysis against another method	`finemap`, `carma`, `rsparsepro`

CLI note

The ABF model itself does not need LD, but the current CLI loader still expects LD files behind each prefix in loci_list.txt. Use the Python API for true no-LD ABF runs.

Single-Input Tools¶

These tools analyze one input at a time. If your loci_list.txt has several rows for the same locus, CREDTOOLS runs the tool per row and combines results.

Tool	Notes
`susie`	strong default; use first
`abf`	simple and can work without LD
`abf_cojo`	ABF with COJO-style conditioning
`finemap`	external FINEMAP engine
`rsparsepro`	alternative sparse model
`carma`	R-based model with outlier handling
`susie_ash`	SuSiE with adaptive shrinkage
`susie_inf`	SuSiE-inf wrapper

Example:

credtools finemap loci_list.txt results_susie --tool susie --max-causal 5

Multi-Input Tools¶

These tools analyze a full locus set together.

Tool	Notes
`multisusie`	good for joint multi-population analysis
`susiex`	designed for cross-population fine-mapping
`mesusie`	multi-ethnic SuSiE-style method

Example:

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

Important Parameters¶

Parameter	Plain meaning
`--max-causal`	maximum number of signals to allow
`--coverage`	credible set coverage, usually `0.95`
`--adaptive-max-causal`	let CREDTOOLS retry with different `max_causal` values
`--set-L-by-cojo`	use COJO to estimate the number of signals
`--purity`	filter credible sets with weak internal LD

A Safe First Run¶

credtools pipeline loci_list.txt results \
  --tool susie \
  --meta-method meta_all \
  --max-causal 5 \
  --coverage 0.95

If many loci hit the maximum number of credible sets, try:

credtools pipeline loci_list.txt results_adaptive \
  --tool susie \
  --adaptive-max-causal

When to Compare Tools¶

Compare tools when:

the credible set drives a downstream experiment,
QC is clean but the credible set is large,
two populations point to different lead variants,
the result is sensitive to --max-causal.

Keep the comparison small. Run several tools on a few important loci before expanding to the full genome.

For exact requirements by tool, see Fine-Mapping Tool Requirements.