mixpred: Combining total and allele-specific reads for building...

In liangyy/mixqtl: Combine total and allele-specific counts for cis-QTL analysis

Description

Given total and allele-specific read counts along with library size, the two-step inference procedure is implemented to build prediction model. Step 1: estimate variances of total and allele-specific count response respectively along with intercept for total count response Step 2: transform observation according to step 1 and build prediction model using glmnet (Elastic net with alpha = alpha and nfold-fold cross-validation)

Usage

mixpred(
  geno1,
  geno2,
  y1,
  y2,
  ytotal,
  lib_size,
  cov_offset = NULL,
  trc_cutoff = 20,
  asc_cutoff = 5,
  weight_cap = 100,
  asc_cap = 5000,
  alpha = 0.5,
  nfold = 5,
  nobs_asc_cutoff = 1
)

Arguments

geno1	genotype of haplotype 1 (dimension = N x P)
geno2	genotype of haplotype 2 (dimension = N x P)
y1	allele-specific read count for haplotype 1 (dimension = N x 1)
y2	allele-specific read count for haplotype 2 (dimension = N x 1)
ytotal	total read count (dimension = N x 1)
lib_size	library size (dimension = N x 1)
cov_offset	predicted effect of covariates on total read count response term, log(ytotal / lib_size / 2) (dimension = N x 1)
trc_cutoff	total read count cutoff to exclude observations with ytotal lower than the cutoff
asc_cutoff	allele-specific read count cutoff to exclude observations with y1 or y2 lower than asc_cutoff
weight_cap	the maximum weight difference (in fold) is min(weight_cap, floor(sample_size / 10)). The ones exceeding the cutoff is capped. Set to NULL then no weight_cap applied.
asc_cap	exclude observations with y1 or y2 higher than asc_cap
alpha	alpha parameter in elastic net model of glmnet (lasso: alpha = 1; ridge: alpha = 0). (default = 0.5).
nfold	number of fold for cross-validation to pick lambda parameter in glmnet. (default = 5).
nobs_asc_cutoff	don't consider ASC if number of observations is smaller than nobs_asc_cutoff

Value

prediction model

Examples

mixpred(
  geno1 = matrix(sample(c(0, 0.5, 1), 200, replace = TRUE), ncol = 2),
  geno2 = matrix(sample(c(0, 0.5, 1), 200, replace = TRUE), ncol = 2),
  y1 = rpois(100, 100),
  y2 = rpois(100, 100),
  ytotal = rpois(100, 1000),
  lib_size = rpois(100, 10000),
  cov_offset = runif(100),
  trc_cutoff = 100,
  asc_cutoff = 50,
  weight_cap = 100,
  asc_cap = 1000
)

liangyy/mixqtl documentation built on Sept. 17, 2020, 11:36 a.m.