postFilterVariants: Post-filtering of Variants

In lawremi/VariantTools: Tools for Exploratory Analysis of Variant Calls

Post-filtering of Variants

Description

Applies filters to a set of called variants. The only current filter is a cutoff on the weighted neighbor count of each variant. This filtering is performed automatically by callVariants, so these functions are for when more control is desired.

Usage

postFilterVariants(x, post.filters = VariantPostFilters(...), ...)
VariantPostFilters(max.nbor.count = 0.1, whitelist = NULL)

Arguments

x	A tally GRanges containing called variants, as output by callVariants.
post.filters	The filters applied to the called variants.
...	Arguments passed to VariantPostFilters, listed below.
max.nbor.count	Maximum allowed number of neighbors (weighted by distance)
whitelist	Positions to ignore; these will always pass the filter, and are excluded from the neighbor counting.

Details

The neighbor count is calculated within a 100bp window centered on the variant. Each neighbor is weighted by the inverse square root of the distance to the neighbor. This was motivated by fitting logistic regression models including a term the count (usually 0, 1, 2) at each distance. The inverse square root function best matched the trend in the coefficients.

Value

For postFilterVariants, a tally GRanges of the variants that pass the filters.

For VariantPostFilters, a FilterRules object with the filters.

Author(s)

Michael Lawrence and Jeremiah Degenhardt

Examples

bams <- LungCancerLines::LungCancerBamFiles()
## post-filters are not enabled by default during calling
data(vignette)
called.variants <- callVariants(tallies_H1993)
## but can be applied at a later time...
postFilterVariants(called.variants, max.nbor.count = 0.15)

# or enable during calling
called.variants <- callVariants(tallies_H1993,
                                post.filters = VariantPostFilters())

lawremi/VariantTools documentation built on March 4, 2024, 11:54 a.m.