binomRegMethModelSim {SOMNiBUS}

R Documentation

Simulate Bisulfite sequencing data from specified smooth covariate effects

Description

Simulate Bisulfite sequencing data from a Generalized Additive Model with functional parameters varying with the genomic position. Both the true methylated counts and observed methylated counts are generated, given the error/conversion rate parameters p0 and p1. In addition, the true methylated counts can be simulated from a binomial or a dispersed binomial distribution (Beta-binomial distribution).

Usage

binomRegMethModelSim(
  n,
  posit,
  theta.0,
  beta,
  phi,
  random.eff = FALSE,
  mu.e = 0,
  sigma.ee = 1,
  p0 = 0.003,
  p1 = 0.9,
  X,
  Z,
  binom.link = "logit"
)

Arguments

n	sample size
posit	genomic position; a numeric vector of size p (the number of CpG sites in the considered region).
theta.0	a functional parameter for the intercept of the GAMM model; a numeric vector of size p.
beta	a functional parameter for the slope of cell type composition. a numeric vector of size p
phi	multiplicative dispersion parameter for each loci in a region. a vector of length p. The dispersed-Binomial counts are simulated from beta-binomial distribution, so each element of phi has to be greater than 1.
random.eff	indicate whether adding the subject-specific random effect term e.
mu.e	the mean of the random effect; a single number.
sigma.ee	variance of the random effect; a single positive number.
p0	the probability of observing a methylated read when the underlying true status is unmethylated. p0 is the rate of false methylation calls, i.e. false positive rate.
p1	the probability of observing a methylated read when the underlying true status is methylated. 1-p1 is the rate of false non-methylation calls, i.e. false negative rate.
X	the matrix of the read coverage for each CpG in each sample; a matrix of n rows and p columns
Z	numeric matrix with p columns and n rows storing the covariate information
binom.link	the link function used for simulation

Value

The function returns a list of following objects

• S the true methylation counts; a numeric matrix of n rows and p columns

• Y the observed methylation counts; a numeric matrix of n rows and p columns

• theta the methylation parameter (after the logit transformation); a numeric matrix of n rows and p columns

• pi the true methylation proportions used to simulate the data; a numeric matrix of n rows and p columns

Author(s)

Kaiqiong Zhao

Examples

#------------------------------------------------------------#
data(RAdat)
RAdat.f <- na.omit(RAdat[RAdat$Total_Counts != 0, ])
out <- binomRegMethModel(
   data=RAdat.f, n.k=rep(5, 3), p0=0, p1=1,
   epsilon=10^(-6), epsilon.lambda=10^(-3), maxStep=200,
   detail=FALSE, RanEff = FALSE
)
Z = as.matrix(RAdat.f[match(unique(RAdat.f$ID), RAdat.f$ID),
c('T_cell', 'RA')])
set.seed(123)
X = matrix(sample(80, nrow(Z)*length(out$uni.pos), replace = TRUE),
nrow = nrow(Z), ncol = length(out$uni.pos))+10
simdat = binomRegMethModelSim(n=nrow(Z), posit= out$uni.pos,
theta.0=out$Beta.out[,1], beta= out$Beta.out[,-1], random.eff=FALSE,
mu.e=0,sigma.ee=1, p0=0.003, p1=0.9,X=X , Z=Z, binom.link='logit',
phi = rep(1, length(out$uni.pos)))

---