In vjcitn/VariantBench:

Introduction

We will illustrate a simple use of the harness for multiple approaches to VCF access. We have a local version of the Tabix-indexed VCF for chr17 for 1000 genomes. We'll set up packages and path.

suppressPackageStartupMessages({
library(VariantBench)
library(GenomicRanges)
})
loc17 = "/Users/stvjc/Research/VCF/ALL.chr17.phase3_shapeit2_mvncall_integrated_v5.20130502.genotypes.vcf.gz"

An illustrative closure is provided that encapsulates the information about the VCF to be processed.

useScanVcfClo

We bind the file path to the VCF processing function.

useScanVcf_local = useScanVcfClo(loc17)
useScanVcf_local 
ls(environment(useScanVcf_local))

Now run the harnessed processing function.

vbHarness(GRanges("17", IRanges(16e6, 16.01e6)), 
    list(useScanVcf=useScanVcf_local))

We can collect information on multiple request types by iterating over ranges of various widths.

rngs = GRanges("17", IRanges(16e6, width=c(1e4, 2e4, 5e4, 1e5)))
multperf = lapply(rngs, function(r)
  vbHarness(r, list(useScanVcf=useScanVcf_local)))

The following overcomplicated functions extract key information about performance. These are fragile to details of the output of the method passed to the harness.

widths = function(x) sapply(x, sapply, 
  function(y) width(y$request))
meantimes = function(y)apply(sapply(y, 
    function(x) (x$useScanVcf$timing$time)), 2, mean)

This permits a plot like

plot( widths(multperf), meantimes(multperf)/10^6, type="b",
  xlab="request width in bp", ylab="time in microsec")

vjcitn/VariantBench documentation built on May 3, 2019, 6:13 p.m.