read_vcf_parallel: Read VCF: parallel

In neurogenomics/MungeSumstats: Standardise summary statistics from GWAS

Read VCF: parallel

Description

Read a VCF file across 1 or more threads in parallel. If tilewidth is not specified, the size of each chunk will be determined by total genome size divided by ntile. By default, ntile is equal to the number of threads, nThread. For further discussion on how this function was optimised, see here and here.

Usage

read_vcf_parallel(
  path,
  samples = 1,
  which = NULL,
  use_params = TRUE,
  as_datatable = TRUE,
  sampled_rows = 10000L,
  include_xy = FALSE,
  download = TRUE,
  vcf_dir = tempdir(),
  download_method = "download.file",
  force_new = FALSE,
  tilewidth = NULL,
  mt_thresh = 100000L,
  nThread = 1,
  ntile = nThread,
  verbose = TRUE
)

Arguments

path	Path to local or remote VCF file.
samples	Which samples to use: 1 : Only the first sample will be used (DEFAULT). NULL : All samples will be used. c("<sample_id1>","<sample_id2>",...) : Only user-selected samples will be used (case-insensitive).
which	Genomic ranges to be added if supplied. Default is NULL.
use_params	When TRUE (default), increases the speed of reading in the VCF by omitting columns that are empty based on the head of the VCF (NAs only). NOTE that that this requires the VCF to be sorted, bgzip-compressed, tabix-indexed, which read_vcf will attempt to do.
as_datatable	Return the data as a data.table (default: TRUE) or a VCF (FALSE).
sampled_rows	First N rows to sample. Set NULL to use full sumstats_file. when determining whether cols are empty.
download	Download the VCF (and its index file) to a temp folder before reading it into R. This is important to keep TRUE when nThread>1 to avoid making too many queries to remote file.
vcf_dir	Where to download the original VCF from Open GWAS. WARNING: This is set to tempdir() by default. This means the raw (pre-formatted) VCFs be deleted upon ending the R session. Change this to keep the raw VCF file on disk (e.g. vcf_dir="./raw_vcf").
download_method	"axel" (multi-threaded) or "download.file" (single-threaded) .
force_new	If a formatted file of the same names as save_path exists, formatting will be skipped and this file will be imported instead (default). Set force_new=TRUE to override this.
tilewidth	The desired tile width. The effective tile width might be slightly different but is guaranteed to never be more than the desired width.
mt_thresh	When the number of rows (variants) in the VCF is < mt_thresh, only use single-threading for reading in the VCF. This is because the overhead of parallelisation outweighs the speed benefits when VCFs are small.
nThread	Number of threads to use for parallel processes.
ntile	The number of tiles to generate.
verbose	Print messages.

Value

VCF file.

Source

path <- "https://gwas.mrcieu.ac.uk/files/ieu-a-298/ieu-a-298.vcf.gz" #### Single-threaded #### vcf <- MungeSumstats:::read_vcf_parallel(path = path) #### Parallel #### vcf2 <- MungeSumstats:::read_vcf_parallel(path = path, nThread=11)

neurogenomics/MungeSumstats documentation built on Aug. 10, 2024, 5:59 a.m.