VariantTools (version 1.14.1)

callVariants: Call Variants

Description

Calls variants from either a BAM file or a VRanges object. The variants are called using a binomial likelihood ratio test. Those calls are then subjected to a post-filtering step.

Usage

"callVariants"(x, tally.param, calling.filters = VariantCallingFilters(...), post.filters = FilterRules(), ...)
"callVariants"(x, ...)
"callVariants"(x, calling.filters = VariantCallingFilters(...), post.filters = FilterRules(), ...)
VariantCallingFilters(read.count = 2L, p.lower = 0.2, p.error = 1/1000)

Arguments

x

Either a path to an indexed bam, a BamFile object, or a VRanges as returned by tallyVariants.

tally.param

Parameters controlling the variant tallying step, as typically constructed by TallyVariantsParam.

calling.filters

Filters used in the calling step, typically constructed with VariantCallingFilters, see arguments listed below.

post.filters

Filters that are applied after the initial calling step. These consider the set of variant calls as a whole and remove those with suspicious patterns.

read.count

Require at least this many high quality reads with the alternate base. The default value is designed to catch sequencing errors where coverage is too low to rely on the LRT. Increasing this value has a significant negative impact on power.

p.lower

The lower bound on the binomial probability for a true variant.

p.error

The binomial probability for a sequencing error (default is reasonable for Illumina data with the default quality cutoff).

...

Arguments to pass to VariantCallingFilters.

Value

For callVariants, a VRanges of the called variants (the tallies that pass the calling filters). See the documentation of bam_tally for complete details.For VariantCallingFilters, a FilterRules object with the filters for calling the variants.

Details

There are two steps for calling variants: the actual statistical test that decides whether a variant exists in the data, and a post-filtering step. By default, the initial calling is based on a binomial likelihood ratio test (P(D|p=p.lower) / P(D|p=p.error) > 1). The test amounts to excluding putative variants with less than ~4% alt frequency. A variant is also required to be represented by at least 2 alt reads. The post-filtering stage considers the set of variant calls as a whole and removes variants with suspicious patterns. Currently, there is a single post-filter, disabled by default, that removes variants that are clumped together on the chromosome (see the max.nbor.count parameter).

Examples

bams <- LungCancerLines::LungCancerBamFiles()
tally.param <- TallyVariantsParam(gmapR::TP53Genome(), 
                                  high_base_quality = 23L,
                                  which = gmapR::TP53Which())

## simple usage
variants <- callVariants(bams$H1993, tally.param)

## customize
calling.filters <- VariantCallingFilters(p.error = 1/1000)
callVariants(bams$H1993, tally.param, calling.filters)