plinkFile (version 0.2.1)

saveGRM: Save symmetic matrix to GCTA GRM format.

Description

GRM (Genetic Relatedness Matrix) is the core formt of GCTA, this function saves a R symmetric matrix to a file set recgnizable by GCTA.

Usage

saveGRM(pfx, grm, vcm = NULL, fid = NULL)

Arguments

pfx

prefix of data files

grm

genome relatedness matrix to save

vcm

variant counts matrix to save (def=1).

fid

separator after family ID. (def=NULL)

Details

Three files will be saved:

.grm.bin :

genetic relatedness matrix in binary

.grm.id :

FID and IID for N individuals in text

.grm.N.bin :

variant count matrix (VCM) in binary

FID and IID will be generated if the grm to be saved has no row names.

When save the vcm, if a single number is given, this number is used as the variant count for all entries in the GRM.

saveGRM is useful in exporting customized kinship matrices (such as a Gaussian or a Laplacian kernel) to a GRM acceptable by GCTA, which are not supported by GCTA's own GRM builder.

Examples

pfx <- file.path(system.file("extdata", package="plinkFile"), "m20")
gmx <- readBED(pfx)  # read genotype matrix from PLINK BED.
gmx <- scale(gmx)    # standardize
tmp <- tempdir()     # for example outputs
dir.create(tmp, FALSE)

# kinship matrix as Gaussian kernel, built from the first 10 variants
gmx.gau <- gmx[, +(1:10)]                 # the first 10 variants
not.na.gau <- tcrossprod(!is.na(gmx.gau)) # variant count matrix
kin.gau <- exp(as.matrix(-dist(gmx.gau, "euc")) / not.na.gau)
print(kin.gau)                            # the Gaussian kernel
out.gau <- file.path(tmp, "m20.gau")
saveGRM(out.gau, kin.gau, not.na.gau)     # gau.grm.* should appear

# kinship matrix as Laplacian kernel, built without the first 10 variants
gmx.lap <- gmx[, -(1:10)]                 # drop the first 10 variants
not.na.lap <- tcrossprod(!is.na(gmx.lap)) # variant count matrix
kin.lap <- exp(as.matrix(-dist(gmx.lap, "man")) / not.na.lap)
out.lap <- file.path(tmp, "m20.lap")
print(kin.lap)                            # the Laplacian kernel
saveGRM(out.lap, kin.lap, not.na.lap)     # lap.grm.* should appear

# merge kinship in R language for a radius based function kernel matrix
not.na.rbf <- not.na.gau + not.na.lap
kin.rbf <- (kin.gau * not.na.gau + kin.lap * not.na.lap) / not.na.rbf
print(kin.rbf)
out.rbf <- file.path(tmp, "m20.rbf")
saveGRM(out.rbf, kin.rbf, not.na.rbf)     # rbf.grm.* should appear

# show saved matrices, then clean up
dir(tmp, "(gau|lap|rbf)")
unlink(tmp, recursive=TRUE)