gnomad_qc.v4.annotations.fix_freq_an
====================================


Updates the v4.0 freq HT with the correct AN for the v4.1 release.

There is an error in the v4.0 freq HT where the AN can be incorrect for variants that
are only present in one of the UKB or non-UKB subsets. This is because the frequency
data was computed on each subset separately and then combined. In order to filter the
dataset to only samples that are in the UKB or non-UKB subsets, the
`hail.vds.filter_samples` was used. Unfortunately, the behavior of this function was
not the expected behavior due to the following line of code:

    .. code-block::

        variant_data = variant_data.filter_rows(hl.agg.count() > 0)

This line of code filters any row that has no genotypes in the sample subset. When the
VariantDataset is densified prior to the frequency calculations this results in no
reference data being filled in for that row, unless a row is part of multiallelic site that is not
exclusive to the sample subset, and therefore a missing AN for that row. When combining
the frequency data for the UKB and non-UKB subsets, the AN for these rows will
only have the AN for the subset that has non-ref genotypes for that row.

This script performs the following steps:

    - Generates allele number and GQ/DP histograms for all sites in the exome calling
      interval.

        - For raw genotypes, the AN is just the sum of the ploidy (more specifics on
          this below) for all genotypes.
        - For adj genotypes, the AN is the sum of the ploidy for all genotypes that pass
          the GQ, DP, and allele balance adj filters. Since this is for a site, rather
          than a specific variant, this number will not always be the same as the AN for
          a specific variant at that site. This AN can be smaller if it is a
          multi-allelic site where there are non-ref genotypes for another variant that
          fails the adj filters.
        - For raw genotypes on non-PAR regions of the X or Y chromosome, the AN is the
          sum of the ploidy after adjusting for sex karyotype, BUT NOT taking into
          account the genotype when adjusting the ploidy. For instance, a het genotype
          for an XY sample will still have a ploidy of 1, even though the genotype is a
          het and is therefore set to missing for the frequency calculations.

    - Generates allele number and GQ/DP histograms for the frequency correction.







Module Functions
****************

.. gnomad_automodulesummary:: gnomad_qc.v4.annotations.fix_freq_an

.. automodule:: gnomad_qc.v4.annotations.fix_freq_an
    :exclude-members: get_script_argument_parser