regenie
Web: https://github.com/rgcgithub/regenie (documentation)
It is the easiest to use the Centos 7 distribution,
export version=v2.2.4
wget -qO- https://github.com/rgcgithub/regenie/releases/download/${version}/regenie_${version}.gz_x86_64_Centos7_mkl.zip | \
gunzip -c > regenie_${version}
chmod +x regenie_${version}
ln -sf regenie_${version} regenie
regenie --help
The last command gives the following information (Why .gz in the banner?),
|=============================|
| REGENIE v2.2.4.gz |
|=============================|
Copyright (c) 2020-2021 Joelle Mbatchou, Andrey Ziyatdinov and Jonathan Marchini.
Distributed under the MIT License.
Compiled with Boost Iostream library.
Using Intel MKL with Eigen.
Usage:
regenie [OPTION...]
-h, --help print list of available options
--helpFull print list of all available options
Main options:
--step INT specify if fitting null model (=1) or
association testing (=2)
--bed PREFIX prefix to PLINK .bed/.bim/.fam files
--pgen PREFIX prefix to PLINK2 .pgen/.pvar/.psam files
--bgen FILE BGEN file
--sample FILE sample file corresponding to BGEN file
--ref-first use the first allele as the reference for
BGEN or PLINK bed/bim/fam input format [default
assumes reference is last]
--keep FILE comma-separated list of files listing samples
to retain in the analysis (no header; starts
with FID IID)
--remove FILE comma-separated list of files listing samples
to remove from the analysis (no header;
starts with FID IID)
--extract FILE comma-separated list of files with IDs of
variants to retain in the analysis
--exclude FILE comma-separated list of files with IDs of
variants to remove from the analysis
-p, --phenoFile FILE phenotype file (header required starting with
FID IID)
--phenoCol STRING phenotype name in header (use for each
phenotype to keep; can use parameter expansion
{i:j})
--phenoColList STRING,..,STRING
comma separated list of phenotype names to
keep (can use parameter expansion {i:j})
-c, --covarFile FILE covariate file (header required starting with
FID IID)
--covarCol STRING covariate name in header (use for each
covariate to keep; can use parameter expansion
{i:j})
--covarColList STRING,..,STRING
comma separated list of covariate names to
keep (can use parameter expansion {i:j})
--catCovarList STRING,..,STRING
comma separated list of categorical
covariates
-o, --out PREFIX prefix for output files
--qt analyze phenotypes as quantitative
--bt analyze phenotypes as binary
-1, --cc12 use control=1,case=2,missing=NA encoding for
binary traits
-b, --bsize INT size of genotype blocks
--cv INT(=5) number of cross validation (CV) folds
--loocv use leave-one out cross validation (LOOCV)
--l0 INT(=5) number of ridge parameters to use when
fitting models within blocks [evenly spaced in
(0,1)]
--l1 INT(=5) number of ridge parameters to use when
fitting model across blocks [evenly spaced in (0,1)]
--lowmem reduce memory usage by writing level 0
predictions to temporary files
--lowmem-prefix PREFIX prefix where to write the temporary files in
step 1 (default is to use prefix from --out)
--split-l0 PREFIX,N split level 0 across N jobs and set prefix of
output files
--run-l0 FILE,K run level 0 for job K in {1..N} using master
file created from '--split-l0'
--run-l1 FILE run level 1 using master file from
'--split-l0'
--keep-l0 avoid deleting the level 0 predictions
written on disk after fitting the level 1 models
--strict remove all samples with missingness at any of
the traits
--print-prs also output polygenic predictions without
using LOCO (=whole genome PRS)
--gz compress output files (gzip format)
--apply-rint apply Rank-Inverse Normal Transformation to
quantitative traits
--threads INT number of threads
--pred FILE file containing the list of predictions files
from step 1
--ignore-pred skip reading predictions from step 1
(equivalent to linear/logistic regression with only
covariates)
--use-prs when using whole genome PRS step 1 output in
'--pred'
--write-samples write IDs of samples included for each trait
(only in step 2)
--minMAC FLOAT(=5) minimum minor allele count (MAC) for tested
variants
--minINFO DOUBLE(=0) minimum imputation info score (Impute/Mach
R^2) for tested variants
--no-split combine asssociation results into a single
for all traits
--firth use Firth correction for p-values less than
threshold
--approx use approximation to Firth correction for
computational speedup
--spa use Saddlepoint approximation (SPA) for
p-values less than threshold
--pThresh FLOAT(=0.05) P-value threshold below which to apply
Firth/SPA correction
--write-null-firth store coefficients from null models with
approximate Firth for step 2
--compute-all store Firth estimates for all chromosomes
--use-null-firth FILE use stored coefficients for null model in
approximate Firth
--chr STRING specify chromosome to test in step 2 (use for
each chromosome)
--chrList STRING,..,STRING
Comma separated list of chromosomes to test
in step 2
--range CHR:MINPOS-MAXPOS
to specify a physical position window for
variants to test in step 2
--sex-specific STRING for sex-specific analyses (male/female)
--af-cc print effect allele frequencies among
cases/controls for step 2
--test STRING 'additive', 'dominant' or 'recessive'
(default is additive test)
--set-list FILE file with sets definition
--extract-sets FILE comma-separated list of files with IDs of
sets to retain in the analysis
--exclude-sets FILE comma-separated list of files with IDs of
sets to remove from the analysis
--extract-setlist STRING comma separated list of sets to retain in the
analysis
--exclude-setlist STRING comma separated list of sets to remove from
the analysis
--anno-file FILE file with variant annotations
--anno-labels FILE file with labels to annotations
--mask-def FILE file with mask definitions
--aaf-file FILE file with AAF to use when building masks
--aaf-bins FLOAT,..,FLOAT
comma separated list of AAF bins cutoffs for
building masks
--build-mask STRING rule to construct masks, can be 'max', 'sum'
or 'comphet' (default is max)
--singleton-carrier define singletons as variants with a single
carrier in the sample
--write-mask write masks in PLINK bed/bim/fam format
--mask-lovo STRING apply Leave-One-Variant-Out (LOVO) scheme
when building masks
(<set_name>,<mask_name>,<aaf_cutoff>)
--mask-lodo STRING apply Leave-One-Domain-Out (LODO) scheme when
building masks
(<set_name>,<mask_name>,<aaf_cutoff>)
--skip-test skip computing association tests after
building masks
--check-burden-files check annotation file, set list file and mask
file for consistency
--strict-check-burden to exit early if the annotation, set list and
mask definition files don't agree
For more information, use option '--help' or visit the website: https://rgcgithub.github.io/regenie/
3.2.7
We could compile from source,
cd ~/rds/public_databases/software/
wget -qO- https://github.com/rgcgithub/regenie/archive/refs/tags/v3.2.7.tar.gz | \
tar xvfz -
cd regenie-3.2.7/
export BGEN_PATH=~/rds/public_databases/software/bgen
module load zlib/1.2.11
export ZLIB_LIBRARY=/usr/local/Cluster-Apps/zlib/1.2.11
module load gcc/6
module load cmake-3.19.7-gcc-5.4-5gbsejo
module load intel/mkl/mic/2018.4
mkdir build
cd build
cmake ..
make
where the bgen and zlib libraries are indicated; module gcc/6 and cmake-3.19.7 are also necessary to get around some other errors.
Note that module load intel/mkl/mic/2018.4 is optional but desirable, and now we have
|============================|
| REGENIE v3.2.7 |
|============================|
Copyright (c) 2020-2022 Joelle Mbatchou, Andrey Ziyatdinov and Jonathan Marchini.
Distributed under the MIT License.
Using Intel MKL with Eigen.
Usage:
./regenie-3.2.7 [OPTION...]
-h, --help print list of available options
--helpFull print list of all available options
Main options:
--step INT specify if fitting null model (=1) or
association testing (=2)
--bed PREFIX prefix to PLINK .bed/.bim/.fam files
--pgen PREFIX prefix to PLINK2 .pgen/.pvar/.psam files
--bgen FILE BGEN file
--sample FILE sample file corresponding to BGEN file
--ref-first use the first allele as the reference for
BGEN or PLINK bed/bim/fam input format
[default assumes reference is last]
--keep FILE comma-separated list of files listing
samples to retain in the analysis (no
header; starts with FID IID)
--remove FILE comma-separated list of files listing
samples to remove from the analysis (no
header; starts with FID IID)
--extract FILE comma-separated list of files with IDs of
variants to retain in the analysis
--exclude FILE comma-separated list of files with IDs of
variants to remove from the analysis
-p, --phenoFile FILE phenotype file (header required starting
with FID IID)
--phenoCol STRING phenotype name in header (use for each
phenotype to keep; can use parameter
expansion {i:j})
--phenoColList STRING,..,STRING
comma separated list of phenotype names to
keep (can use parameter expansion {i:j})
-c, --covarFile FILE covariate file (header required starting
with FID IID)
--covarCol STRING covariate name in header (use for each
covariate to keep; can use parameter
expansion {i:j})
--covarColList STRING,..,STRING
comma separated list of covariate names to
keep (can use parameter expansion {i:j})
--catCovarList STRING,..,STRING
comma separated list of categorical
covariates
-o, --out PREFIX prefix for output files
--qt analyze phenotypes as quantitative
--bt analyze phenotypes as binary
-1, --cc12 use control=1,case=2,missing=NA encoding
for binary traits
-b, --bsize INT size of genotype blocks
--cv INT(=5) number of cross validation (CV) folds
--loocv use leave-one out cross validation (LOOCV)
--l0 INT(=5) number of ridge parameters to use when
fitting models within blocks [evenly spaced
in (0,1)]
--l1 INT(=5) number of ridge parameters to use when
fitting model across blocks [evenly spaced
in (0,1)]
--lowmem reduce memory usage by writing level 0
predictions to temporary files
--lowmem-prefix PREFIX prefix where to write the temporary files
in step 1 (default is to use prefix from
--out)
--split-l0 PREFIX,N split level 0 across N jobs and set prefix
of output files
--run-l0 FILE,K run level 0 for job K in {1..N} using
master file created from '--split-l0'
--run-l1 FILE run level 1 using master file from
'--split-l0'
--l1-phenoList STRING,...,STRING
run level 1 for a subset of the phenotypes
(specified as comma-separated list)
--keep-l0 avoid deleting the level 0 predictions
written on disk after fitting the level 1
models
--strict remove all samples with missingness at any
of the traits
--print-prs also output polygenic predictions without
using LOCO (=whole genome PRS)
--gz compress output files (gzip format)
--apply-rint apply Rank-Inverse Normal Transformation to
quantitative traits
--apply-rerint apply Rank-Inverse Normal Transformation to
residualized quantitative traits in step 2
--apply-rerint-cov apply Rank-Inverse Normal Transformation to
residualized quantitative traits and
project covariates out in step 2
--threads INT number of threads
--pred FILE file containing the list of predictions
files from step 1
--ignore-pred skip reading predictions from step 1
(equivalent to linear/logistic regression
with only covariates)
--use-prs when using whole genome PRS step 1 output
in '--pred'
--write-samples write IDs of samples included for each
trait (only in step 2)
--minMAC FLOAT(=5) minimum minor allele count (MAC) for tested
variants
--minINFO DOUBLE(=0) minimum imputation info score (Impute/Mach
R^2) for tested variants
--no-split combine asssociation results into a single
for all traits
--firth use Firth correction for p-values less than
threshold
--approx use approximation to Firth correction for
computational speedup
--spa use Saddlepoint approximation (SPA) for
p-values less than threshold
--pThresh FLOAT(=0.05) P-value threshold below which to apply
Firth/SPA correction
--write-null-firth store coefficients from null models with
approximate Firth for step 2
--compute-all store Firth estimates for all chromosomes
--use-null-firth FILE use stored coefficients for null model in
approximate Firth
--chr STRING specify chromosome to test in step 2 (use
for each chromosome)
--chrList STRING,..,STRING
Comma separated list of chromosomes to test
in step 2
--range CHR:MINPOS-MAXPOS
to specify a physical position window for
variants to test in step 2
--sex-specific STRING for sex-specific analyses (male/female)
--af-cc print effect allele frequencies among
cases/controls for step 2
--test STRING 'additive', 'dominant' or 'recessive'
(default is additive test)
--condition-list FILE file with list of variants to include as
covariates
--condition-file FORMAT,FILE
optional genotype file which contains the
variants to include as covariates
--condition-file-sample FILE
sample file accompanying BGEN file with the
conditional variants
--interaction STRING perform interaction testing with a
quantitative/categorical covariate
--interaction-snp STRING perform interaction testing with a variant
--interaction-file FORMAT,FILE
optional genotype file which contains the
variant for GxG interaction test
--interaction-file-sample FILE
sample file accompanying BGEN file with the
interacting variant
--interaction-file-reffirst
use the first allele as the reference for
the BGEN or PLINK file with the interacting
variant [default assumes reference is last]
--interaction-prs perform interaction testing with the full
PRS from step 1
--force-condtl to also condition on interacting SNP in the
marginal GWAS test
--no-condtl to print out all main effects in GxE
interaction test
--rare-mac FLOAT(=1000) minor allele count (MAC) threshold below
which to use HLM for interaction testing
with QTs
--set-list FILE file with sets definition
--extract-sets FILE comma-separated list of files with IDs of
sets to retain in the analysis
--exclude-sets FILE comma-separated list of files with IDs of
sets to remove from the analysis
--extract-setlist STRING comma separated list of sets to retain in
the analysis
--exclude-setlist STRING comma separated list of sets to remove from
the analysis
--anno-file FILE file with variant annotations
--anno-labels FILE file with labels to annotations
--mask-def FILE file with mask definitions
--aaf-file FILE file with AAF to use when building masks
--set-singletons use 0/1 indicator in third column of AAF
file to specify singleton variants
--aaf-bins FLOAT,..,FLOAT
comma separated list of AAF bins cutoffs
for building masks
--build-mask STRING rule to construct masks, can be 'max',
'sum' or 'comphet' (default is max)
--vc-tests STRING,..,STRING
comma separated list of tests to compute
for each set of variants included in a mask
[skat/skato/skato-acat/acatv/acato]
--vc-maxAAF FLOAT(=1) maximum AAF for variants included in
gene-based tests
--weights-col arg column index (1-based) for user-defined
weights in annotation file
--multiply-weights multiply the user defined weights by the
default SKAT weights in SKAT/ACAT tests
--joint STRING comma spearated list of joint tests to
perform
--singleton-carrier define singletons as variants with a single
carrier in the sample
--write-mask write masks in PLINK bed/bim/fam format
--mask-lovo STRING apply Leave-One-Variant-Out (LOVO) scheme
when building masks
(<set_name>,<mask_name>,<aaf_cutoff>)
--mask-lodo STRING apply Leave-One-Domain-Out (LODO) scheme
when building masks
(<set_name>,<mask_name>,<aaf_cutoff>)
--skip-test skip computing association tests after
building masks
--check-burden-files check annotation file, set list file and
mask file for consistency
--strict-check-burden to exit early if the annotation, set list
and mask definition files don't agree
--force-qt force QT run for traits with few unique
values
--par-region STRING(=hg38)
build code to identify PAR region
boundaries on chrX
For more information, use option '--help' or visit the website: https://rgcgithub.github.io/regenie/
Reference
Mbatchou, J., Barnard, L., Backman, J. et al. Computationally efficient whole-genome regression for quantitative and binary traits. Nat Genet 53, 1097–1103 (2021). https://doi.org/10.1038/s41588-021-00870-7