Skip to content

usage-guide.md

Latest commit

 

History

History
78 lines (52 loc) · 2.58 KB

usage-guide.md

File metadata and controls

78 lines (52 loc) · 2.58 KB

Fine-Mapping - Usage Guide

Overview

Identify likely causal variants within GWAS loci by computing posterior inclusion probabilities (PIPs) and constructing credible sets. SuSiE handles multiple causal variants and integrates with colocalization analysis.

Prerequisites

install.packages('susieR')
install.packages(c('ggplot2', 'patchwork'))

# FINEMAP (command-line tool)
# Download from http://www.christianbenner.com/
# For LD matrix generation
# plink v1.9+ required
conda install -c bioconda plink

Quick Start

Tell your AI agent what you want to do:

  • "Fine-map this GWAS locus to a 95% credible set"
  • "Compute PIPs for all variants at my GWAS hit using SuSiE"
  • "Run FINEMAP on this locus with the LD matrix from 1000 Genomes"

Example Prompts

SuSiE Fine-Mapping

"I have GWAS summary stats and an LD matrix for a locus on chromosome 6 -- run SuSiE and report the credible sets"

"Fine-map my GWAS signal using susie_rss with L = 10"

FINEMAP

"Set up FINEMAP input files for this locus and run the stochastic search"

"Compare SuSiE and FINEMAP results at my top GWAS locus"

Visualization

"Make a PIP plot alongside the GWAS Manhattan for this locus"

"Color the regional plot by credible set membership"

LD Reference

"Generate an LD matrix from 1000 Genomes Europeans for this 1 Mb region"

What the Agent Will Do

  1. Extract locus data (typically 1 Mb window around lead SNP)
  2. Obtain or compute LD matrix from matched ancestry reference
  3. Run SuSiE (susie_rss) with summary statistics and LD
  4. Extract 95% credible sets and PIPs
  5. Report credible set size, purity, and top PIP variants
  6. Optionally run FINEMAP for comparison
  7. Generate PIP and regional association plots

Tips

  • L parameter - Start with L = 10; SuSiE automatically prunes unused effects
  • LD ancestry match - The LD reference must match the GWAS sample ancestry; mismatched LD causes false signals
  • Credible set purity - Minimum absolute correlation > 0.5 indicates a well-resolved signal
  • PIP thresholds - PIP > 0.95 is strong evidence; PIP > 0.5 is suggestive
  • FINEMAP comparison - When SuSiE and FINEMAP agree, results are more reliable
  • Positive semi-definite - Add a small ridge to the LD matrix diagonal if eigenvalues are negative

Related Skills

  • causal-genomics/colocalization-analysis - SuSiE-coloc for shared causal variants
  • causal-genomics/mendelian-randomization - Fine-map instrument loci
  • population-genetics/linkage-disequilibrium - LD reference panels
  • variant-calling/variant-annotation - Annotate fine-mapped variants