RareEffect

RareEffect is a novel method to estimate the variant-level effect size (beta) and region-level (gene-level) heritability for rare variants. RareEffect is integrated into SAIGE and is available in SAIGE version 1.3.x or higher.

Procedure

  1. First, run SAIGE step 1. The output file from SAIGE step 1 (.rda file) will be used as the input for RareEffect.
  2. The inputs for RareEffect are as follows:
  • Genotype file (plink bed/bim/fam)
  • Group file (same format as the group file for SAIGE-GENE+)
  • SAIGE step 1 output rda file
  1. Execute the main function of RareEffect. An example command is provided below.
Rscript run_RareEffect.R \
    --rdaFile=$PATH_TO_RDA_FILE \
    --chrom=$CHR \
    --geneName=$GENE_NAME \
    --groupFile=$PATH_TO_GROUP_FILE \
    --traitType=binary \
    --bedFile=$PATH_TO_BED_FILE \
    --bimFile=$PATH_TO_BIM_FILE \
    --famFile=$PATH_TO_FAM_FILE \
    --macThreshold=10 \
    --collapseLoF=FALSE \
    --outputPrefix=$PATH_TO_OUTPUT_FILE
  • rdaFile: SAIGE step 1 output file
  • chrom: chromosome number (eg. 11)
  • geneName: name of gene to analyze (eg. APOC3)
  • groupFile: group file containing variant list and functional annotations.
  • traitType: type of trait to be analyzed, binary or quantitative
  • bedFile, bimFile, famFile: genotype file
  • macThreshold: minor allele count threshold for ultra-rare variant collapsing
  • collapseLoF: if true, RareEffect collapses all LoF variants into one super-variant like Burden test (regardless of their MAC)
  • outputPrefix: path to output
  1. RareEffect generates two output files: variant-level effect size and region-level heritability.
  • Variant-level effect size
variant          effect              PEV
11:116830637:C:T 0.304491201231766   0.00727297925199994
11:116830638:G:A 0.325660176157425   0.000662034527515423
11:116830638:G:T 0.21230324192781    0.0645359637154959
11:116830897:G:T 0.310004879463224   0.060623563613811
lof_UR           0.261071923418984   0.0869925579756295
11:116830533:G:A -0.0631819446554343 0.0235133013294374
11:116830620:C:T -0.0130476721146154 0.0539691054813582
11:116830787:G:A 0.0448881905903484  0.0317167120002997
11:116830833:C:A 0.310894962948392   0.0570479313033626
  • Region-level heritability
LoF                 mis                 syn                  all
0.00326766343863732 0.00028977664861778 3.21446473889367e-05 0.00358958473464404
  1. Using the variant-level effect size from RareEffect, you can calculate the polygenic score in another cohort.
Rscript calculate_RareEffect_PRS.R \
    --effectFile $PATH_TO_EFFECT_SIZE_FILE \
    --bedFile $PATH_TO_BED_FILE \
    --bimFile $PATH_TO_BIM_FILE \
    --famFile $PATH_TO_FAM_FILE \
    --groupFile $PATH_TO_GROUP_FILE \
    --geneName $GENE_NAME \
    --variantListFile $PATH_TO_VARIANT_LIST_FILE \
    --outputPrefix $PATH_TO_OUTPUT_FILE
  • effectFile: variant-level effect size output from RareEffect main function

  • variantListFile: list of common and rare, but non-ultra-rare variants. These variants will not be used for ultra-rare variant scoring.

  • Individual RareEffect PRS 
    IID     PRS
1000019 0
1000022 0
1000035 0
1000046 0
1000054 0
1000063 0
1000078 0.48012911
1000081 0
1000090 0
1000105 0

Example output

We estimated variant-level effect size and gene-level heritability for genome-wide significant genes (or top 10 genes) using 393,247 White British individuals in UKB WES data. The example output files can be downloaded at:

  • Variant-level effect size: https://storage.googleapis.com/leelabsg/RareEffect/RareEffect_effect_size.zip
  • Gene-level signed heritability: https://storage.googleapis.com/leelabsg/RareEffect/RareEffect_h2.zip