ctcisQTL {omicwas}

Cell-Type-Specific QTL analysis

Description

Cell-Type-Specific QTL analysis

Usage

ctcisQTL(
  X,
  Xpos,
  W,
  Y,
  Ypos,
  C = NULL,
  max.pos.diff = 1e+06,
  outdir = tempdir(),
  outfile = "ctcisQTL.out.txt"
)

Arguments

X

Matrix (or vector) of SNP genotypes; SNPs x samples.

Xpos

Vector of the physical position of X

W

Matrix of cell type composition; samples x cell types.

Y

Matrix (or vector) of bulk omics measurements; markers x samples.

Ypos

Vector of the physical position of Y

C

Matrix (or vector) of covariates; samples x covariates. X, Xpos, W, Y, Ypos, C should be numeric.

max.pos.diff

Maximum positional difference to compute cis-QTL. Association analysis is performed between a row of X and a row of Y, only when they are within this limit. Since the limiting is only by position, the function needs to be run separately for each chromosome.

outdir

Output directory.

outfile

Output file.

Details

A function for analyses of QTL, such as eQTL, mQTL, pQTL. The statistical test is almost identical to ctassoc(test = "nls.identity", regularize = "TRUE"). Association analysis is performed between each row of Y and each row of X. Usually, the former will be a methylation/expression marker, and the latter will be a SNP. To cope with the large number of combinations, the testing is limited to pairs whose position is within the difference specified by max.pos.diff; i.e., limited to cis-QTL. In detail, this function performs linear ridge regression, whereas ctassoc(test = "nls.identity", regularize = "TRUE") actually is nonlinear regression but with f = identity as normalizing transformation. In order to speed up computation, first, the parameters \alpha_{h j} and \gamma_{j l} are fit by ordinary linear regression assuming \beta_{h j k} = 0. Next, \beta_{h j k} are fit and tested by linear ridge regression (see documentation for ctassoc).

Value

The estimate, statistic, p.value are written to the specified file.

See Also

ctassoc

Examples

data(GSE79262small)
X    = GSE79262small$X
Xpos = GSE79262small$Xpos
W    = GSE79262small$W
Y    = GSE79262small$Y
Ypos = GSE79262small$Ypos
C    = GSE79262small$C
X    = X[seq(1, 3601, 100), ] # for brevity
Xpos = Xpos[seq(1, 3601, 100)]
ctcisQTL(X, Xpos, W, Y, Ypos, C = C)