VCF parsing

VCF parsing#

fastdfe infers a DFE from a site-frequency spectrum, and everything that turns a VCF into one lives in sfsutils, the standalone package fastdfe depends on and re-exports. Every class below is therefore available under both names, and resolves to the same sfsutils class.

Note

As of fastdfe 1.4.0, code written against earlier releases remains valid without modification. See the changelog for the details of the separation.

import fastdfe as fd
import sfsutils as su

fd.Parser.__module__, su.Parser.__module__

('sfsutils.parser', 'sfsutils.parser')

library(fastdfe)

fd <- load_fastdfe()
su <- reticulate::import("sfsutils")

c(fd$Parser$`__module__`, su$Parser$`__module__`)

[1] "sfsutils.parser" "sfsutils.parser"

Parser reads a VCF and returns the Spectra that the inference pages take as input. It determines polarization from the AA tag by default (see skip_non_polarized), and takes three kinds of component:

  • Stratifications split the spectrum by site property, most importantly DegeneracyStratification, which gives the neutral and selected spectra a DFE inference needs.

  • Annotations add the site information a stratification reads. DegeneracyAnnotation determines each site’s degeneracy from a reference genome and a GFF, which DegeneracyStratification then splits on, and an ancestral allele annotation provides the polarization an unfolded spectrum requires.

  • Filtrations drop sites before they reach the spectrum, such as BiasedGCConversionFiltration, and are applied to a VCF directly with the Filterer.

For worked examples in Python and R, see sfsutils’ guides on parsing, annotations, filtrations and manipulating the SFS.