GenABEL (version 1.8-0)

blurGenotype: blur genotype calls into probabilites

Description

'blurs' genotype calls into probabilities: translates single genotype g2, into probability distribution P(g1|g2), that is probability that true genotype is g1 given g2 is the observed 'called' genotype and error rate is epsilon. Probability that 'true' genotype is called genotype is set to (1-epsilon)^2, the probability that true genotype differs at 1 allele is set to epsilon*(1-epsilon), and both allelel differ = epsilon^2.

Usage

blurGenotype(g, q = NULL, epsilon = 0.01)

Arguments

g

vector of genotypes for a particular person (at locus 1, locus 2, etc., coded as 0, 1, 2 (corresponding to genotypes AA, AB, and BB, respectively) and NA.

q

(optional) vector of coded allele freqeuncies for locus 1, locus 2, etc.

epsilon

error rate

Value

matrix with columns corresponding to SNPs and rows corresponding to 'g0', 'g1', 'g2'. For a particular SNP, a vale in cell 'gK' is the probability that true genotype is 'K', given thw original call and error-rate.

Examples

require(GenABEL.data)
data(srdta)
# select 10 first SNPs
df <- srdta[,1:10]
# compute effect allele freq
EAF <- summary(gtdata(df))$"Q.2"
EAF
# get genotypes of first 5 people
g1 <- as.numeric(df[1:5,])
g1
# blur the genotype of person 1, snp 1
blurGenotype(g1[1,1])
# blur all genotypes of person 2; assume no info for missing
blurGenotype(g1[2,])
# blur all genotypes of person 2; use HWE to infer missing
blurGenotype(g1[2,],q=EAF)