sequoia (version 2.5.6)

MkGenoErrors: Simulate Genotyping Errors

Description

Generate errors and missing values in a (simulated) genotype matrix.

Usage

MkGenoErrors(
  SGeno,
  CallRate = 0.99,
  SnpError = 5e-04,
  ErrorFM = function(E) {
     matrix(c(1 - E - (E/2)^2, E, (E/2)^2, E/2, 1 - E, E/2,
    (E/2)^2, E, 1 - E - (E/2)^2), 3, 3, byrow = TRUE)
 },
  Error.shape = 0.5,
  CallRate.shape = 1
)

Value

The input genotype matrix, with some genotypes replaced, and some set to missing (-9).

Arguments

SGeno

matrix with genotype data in Sequoia's format: 1 row per individual, 1 column per SNP, and genotypes coded as 0/1/2.

CallRate

either a single number for the mean call rate (genotyping success), OR a vector with the call rate at each SNP, OR a named vector with the call rate for each individual. In the third case, ParMis is ignored, and individuals in the pedigree (as id or parent) not included in this vector are presumed non-genotyped.

SnpError

mean per-locus genotyping error rate across SNPs, and a beta-distribution will be used to simulate the number of missing cases per SNP, OR a vector with the genotyping error for each SNP.

ErrorFM

function taking the error rate (scalar) as argument and returning a 4x4 or 3x3 matrix with probabilities that actual genotype i (rows) is observed as genotype j (columns).

Error.shape

first shape parameter (alpha) of beta-distribution of per-SNP error rates. A higher value results in a flatter distribution.

CallRate.shape

as Error.shape, for per-SNP call rates.

Examples

GenoM <- SimGeno(Ped = Ped_HSg5, nSnp = 100, ParMis = 0.2,
                 SnpError=0, CallRate=1)
GenoM.actual <- GenoM
LowQ <- sample.int(nrow(GenoM), 42)  # low-quality samples
GenoM[LowQ, ] <- MkGenoErrors(GenoM[LowQ, ], SnpError = 0.05)
GenoM[-LowQ, ] <- MkGenoErrors(GenoM[-LowQ, ], SnpError = 0.001)
ErrorCount <- sapply(1:nrow(GenoM), function(i) {
  sum(GenoM.actual[i,] != GenoM[i,] & GenoM[i,] != -9) } )
mean(ErrorCount[LowQ])
mean(ErrorCount[-LowQ])