check.marker: function to do genotypic quality control

Description

This function helps selecting the marker which should enter into GWA analysis based on call rate, minor allele frequency, value of the chi-square test for Hardy-Weinberg equilibrium, and redudndance, defined as concordance between the distributions of the genotypes (including missing values).

Usage

check.marker(data, snpsubset, idsubset, callrate = 0.95,  perid.call=0.95, extr.call = 0.1, extr.perid.call = 0.1, het.fdr = 0.01,  ibs.threshold = 0.95, ibs.mrk = 2000, ibs.exclude="both", maf, p.level = -1,  fdrate = 0.2, odds = 1000, hweidsubset, redundant = "no",  minconcordance = 2.0, qoption = "bh95", imphetasmissing = TRUE, XXY.call=0.8, intermediateXF = c(0.5, 0.5))

Arguments

data

gwaa.data or snp.data object

snpsubset

a subset of SNPs to check (names, indexes, logical), default is all from data

idsubset

a subset of people to check (names, indexes, logical), default is all from data

callrate

cut-off SNP call rate

perid.call

cut-off individual call rate (maximum percent of missing genotypes in a person)

extr.call

SNPs with this low call rate are dropped prior to main analysis

extr.perid.call

people with this low call rate are dropped prior to main analysis

het.fdr

FDR rate for unacceptably high individual heterozygosity

ibs.threshold

threshold value for acceptable IBS

ibs.mrk

How many random markers should be used to estimate IBS. When ibs.mrk < 1, IBS checks are turned off. When "all" all markers are used.

ibs.exclude

"both", "lower" or "none" -- whether both samples with IBS>ibs.threshold should be excluded, the one with lower call rate, or no check (equivalent to use of 'ibs.mrk = -1').

maf

cut-off Minor Allele Frequency. If not specified, the default value is 5 chromosomes 5/(2*nids(data))

p.level

cut-off p-value in check for Hardy-Weinberg Equilibrium. If negative, FDR is applied

fdrate

cut-off FDR level in check for Hardy-Weinberg Equilibrium

odds

cut-off odds to decide whether marker/person should be excluded based on sex/X-linked marker data inconsistency

hweidsubset

a subset of people to check (names, indexes, logical) to use for HWE check

redundant

if "bychrom", redundancy is checked within chromosomes; "all" -- all pairs of markers; "no" -- no redundancy checks

minconcordance

a parameter passed to "redundant" function. If "minconcordance" is > 1.0 only pairs of markers which are exactly the same, including NA pattern, are considered as redundant; if 0 < "minconcordance" < 1, then pairs of markers with concordance > "minconcordance" are considered redundant. see redundant for details. Note that if "minconcordance" < 1 the program will take much longer time to run

qoption

if "bh95", BH95 FDR used; if "storey", qvalue package (if installed) is used

imphetasmissing

If "impossible heterozygotes" (e.g. heterozygous mtDNA, and male Y- and X-chromsome markers) should be treated as missing genotypes in the QC procedure

XXY.call

What proportion of Y-chromosome markers should be called to consider that Y-chromosome is present (in presence of XX)

intermediateXF

X-chromosomal F to be considered 'intermediate' and regarded as error; currently use of default disables this check

Value

Object of class check.marker-class

Details

In this procedure, sex errors are identified initally and then possible residual errors are removed iteratively. At the first step, of the iterative procedure, per-marker (minor allele frequency, call rate, exact P-value for Hardy-Weinberg equilibrium) and between-marker statistics are generated and controlled for, mostly using the internal call to the function summary.snp.data. At the second step of the iterative procedure, per-person statistics, such call rate within a person, heterozygosity and and between-person statistics (identity by state across a random sample of markers) are generated, using perid.summary and ibs functions. Heterozygosity and IBS are estimated using only autosomal data. If IBS is over ibs.threshold for a pair, one person from the pair is added to the ibsfail list and excluded from the idok list. At the second step, only the markers passing the first step are used.

The procedure is applied recursively till no further markers and people are eliminated.

Examples

Run this code

# usual way
require(GenABEL.data)
data(ge03d2c)
# truncate the data to make the example faster
ge03d2c <- ge03d2c[seq(from=1,to=nids(ge03d2c),by=2),seq(from=1,to=nsnps(ge03d2c),by=2)]
# many errors
mc0 <- check.marker(ge03d2c)
# take only people and markers passing QC
fixed0 <- ge03d2c[mc0$idok,mc0$snpok]
# major errors fixed, still few males are heterozygous for X-chromsome markers
mc1 <- check.marker(fixed0)
# fix minor X-chromosome problems
fixed1 <- Xfix(fixed0)
# no errors
mc2 <- check.marker(fixed1)
summary(mc2)
# ready to use fixed1 for analysis

# let us look into redundancy
require(GenABEL.data)
data(srdta)
mc <- check.marker(data=srdta,ids=c(1:300),call=.92,perid.call=.92)
names(mc)
mc$nohwe
mc <- check.marker(data=srdta@gtdata[,1:100],call=0.95,perid.call=0.9,
		maf=0.02,minconcordance=0.9,fdr=0.1,redundant="all",ibs.mrk=0)
summary(mc)
HWE.show(data=srdta,snps=mc$nohwe)
plot(mc)

Run the code above in your browser using DataLab

Get 50% off unlimited learning