Check a long-format sequence-derived marker genotype file

Validates the structure of a long-format marker genotype table (one row per id x locus), the same schema checkMarkerGenotypeFile checks – but sized and hardened for sequence-derived panels (issue #152): a soft, overridable warning above a sparse/GBS-scale panel-size ceiling, and an explicit rejection of a literal "." allele value (VCF's missing-genotype placeholder), rather than silently counting it as a real allele. Optionally cross-validates an accompanying locus-metadata sidecar by reusing checkLocusMetadata, rather than reinventing that check.

Usage

checkSequenceGenotypeFile(genotype, locusMetadata = NULL, maxLoci = 50000L)

Arguments

genotype: dataframe with long-format marker genotype data: exactly four columns, id, locus, allele1, allele2 (one row per individual x locus).
locusMetadata: optional dataframe with locus metadata (see checkLocusMetadata): locus, chrom, pos, and optionally cM, one row per locus. When supplied, is validated via checkLocusMetadata – its own violations propagate unchanged. Defaults to NULL (no sidecar to validate).
maxLoci: numeric scope-tier ceiling (default 50000L, this package's own sparse/GBS-scale target ceiling) above which a locus count triggers a warning() rather than a stop().

Value

The genotype dataframe, checked to ensure the column count, first-column identity, per-locus allele count, absence of a literal "." placeholder, and row uniqueness are all valid. The returned dataframe has its column names forced to c("id", "locus", "allele1", "allele2").

Details

All of checkMarkerGenotypeFile's structural checks are retained – exactly four columns, id as the first column, no duplicate id x locus rows, and the same biallelic-only rejection the KING-robust kinship estimator (Manichaikul et al. 2010) requires. Two rules are new: a literal "." allele value is rejected before the biallelic count is even checked, so a curator sees the correct, actionable error rather than a misleading "more than two alleles" report (a genotype-preprocessing pipeline emitting VCF-style missingness that was never converted to NA is a real, anticipated failure mode – see Danecek et al. 2011 for the VCF missing-genotype convention this guards against); and a locus count above maxLoci produces a warning(), not a stop() – the "right" ceiling for this package's own vectorized implementations is not yet empirically known, so a hard stop would risk blocking legitimate data.

References

Manichaikul, A., Mychaleckyj, J. C., Rich, S. S., Daly, K., Sale, M., & Chen, W.-M. (2010). Robust relationship inference in genome-wide association studies. Bioinformatics, 26(22), 2867-2873. doi:10.1093/bioinformatics/btq559

Danecek, P., et al. (2011). The variant call format and VCFtools. Bioinformatics, 27(15), 2156-2158. doi:10.1093/bioinformatics/btr330

Examples

library(nprcgenekeepr)
markerGenotype <- data.frame(
  id = c("A", "A", "B", "B"),
  locus = c("L1", "L2", "L1", "L2"),
  allele1 = c("A", "A", "A", "A"),
  allele2 = c("A", "B", "B", "B"),
  stringsAsFactors = FALSE
)
checkSequenceGenotypeFile(markerGenotype)
#>   id locus allele1 allele2
#> 1  A    L1       A       A
#> 2  A    L2       A       B
#> 3  B    L1       A       B
#> 4  B    L2       A       B