A set-based association test in the snpsettest package requires a reference data set to infer pairwise linkage disequilibrium (LD) values between a set of variants. This vignette shows you how to use 1000 Genomes data as the reference data for set-based association tests.
PLINK 2.0 is required to process the 1000 Genomes dataset. The 1000 Genomes phase 3 dataset (GRCh37) is available in PLINK2 binary format at PLINK 2.0 Resources. To download files,
# The links in here may be changed in future
# "-O" to specify output file name
wget -O all_phase3.psam "https://www.dropbox.com/s/yozrzsdrwqej63q/phase3_corrected.psam?dl=1"
wget -O all_phase3.pgen.zst "https://www.dropbox.com/s/afvvf1e15gqzsqo/all_phase3.pgen.zst?dl=1"
wget -O all_phase3.pvar.zst "https://www.dropbox.com/s/op9osq6luy3pjg8/all_phase3.pvar.zst?dl=1"
# Decompress pgen.zst to pgen
plink2 --zst-decompress all_phase3.pgen.zst > all_phase3.pgen
Patterns of LD could vary among racial/ethnic groups, and thus, it may be necessary to choose an appropriate population. For example, if your GWAS is based on European descent, you may want to keep EUR samples as described in the “all_phase3.psam” file.
# "vzs" modifier to directly operate with pvar.zst
# "--chr 1-22" excludes all variants not on the listed chromosomes
# "--output-chr 26" uses numeric chromosome codes
# "--max-alleles 2": PLINK 1 binary does not allow multi-allelic variants
# "--rm-dup" removes duplicate-ID variants
# "--set-missing-var-id" replaces missing IDs with a pattern
plink2 --pfile all_phase3 vzs \
--chr 1-22 \
--output-chr 26 \
--max-alleles 2 \
--rm-dup exclude-mismatch \
--set-missing-var-ids '@_#_$1_$2' \
--make-pgen \
--out all_phase3_autosomes
# Prepare sub-population filter file
awk 'NR == 1 || $5 == "EUR" {print $1}' all_phase3.psam > EUR_1kg_samples.txt
# Generate sub-population fileset
plink2 --pfile all_phase3_autosomes \
--keep EUR_1kg_samples.txt \
--make-pgen \
--out EUR_phase3_autosomes
The snpsettest package uses PLINK 1 binary files to read them into R. The PLINK2 binary fileset (pgen/pvar/psam) can be easily converted to PLINK 1 binary fileset (bed/bim/fam).
# pgen to bed
# "--maf 0.005" remove most monomorphic SNPs
# (still may have some when all samples are heterozyguous -> maf=0.5)
plink2 --pfile EUR_phase3_autosomes \
--maf 0.005 \
--make-bed \
--out EUR_phase3_autosomes
# Split bed/bim/fam by chromosome
for i in {1..22}
do plink2 --bfile EUR_phase3_autosomes --chr $i --make-bed --out EUR_phase3_chr$i
done
For the snpsettest package, it is better to split your reference data by chromosome and run set-based association tests with per-chromosome binary files. For instance, when you perform set-based association tests for genes on chromosome 1, you don’t have to load genotype data for other chromosomes into R. Intuitively, as more redundant SNPs are included in the reference data, your tests will get (often significantly) slower and consume more memory.
This package uses a bed.matrix-class from the gaston package to attach genotype data to R session. Genotypes are retrieved on demand to manage large-scale genotype data.