callMethylation: Call methylation status

In methimpute: Imputation-guided re-construction of complete methylomes from WGBS data

Description

Call methylation status of cytosines (or bins) with a Hidden Markov Model.

Usage

callMethylation(data, fit.on.chrom = NULL, transDist = Inf, eps = 1,
  max.time = Inf, max.iter = Inf, count.cutoff = 500,
  verbosity = 1, num.threads = 2 + include.intermediate,
  initial.params = NULL, include.intermediate = FALSE,
  update = "independent", min.reads = 0)

Arguments

data	A methimputeData object.
fit.on.chrom	A character vector specifying the chromosomes on which the HMM will be fitted.
transDist	The decaying constant for the distance-dependent transition matrix. Either a single numeric or a named numeric vector, where the vector names correspond to the transition contexts. Such a vector can be obtained from estimateTransDist.
eps	Convergence threshold for the Baum-Welch algorithm.
max.time	Maximum running time in seconds for the Baum-Welch algorithm.
max.iter	Maximum number of iterations for the Baum-Welch algorithm.
count.cutoff	A cutoff for the counts to remove artificially high counts from mapping artifacts. Set to Inf to disable this filtering (not recommended).
verbosity	An integer from 1 to 5 specifying the verbosity of the fitting procedure. Values > 1 are only for debugging.
num.threads	Number of CPU to use for the computation. Parallelization is implemented on the number of states, which is 2 or 3, so setting num.threads > 3 will not give additional performance increase.
initial.params	A methimputeBinomialHMM object. This parameter is useful to continue the fitting procedure for a methimputeBinomialHMM object.
include.intermediate	A logical specifying wheter or not the intermediate component should be included in the HMM.
update	One of c("independent", "constrained"). If update="independent" probability parameters for the binomial test will be updated independently. If update="constrained" the probability parameter of the intermediate component will be constrained to the mean of the unmethylated and the methylated component.
min.reads	The minimum number of reads that a position must have to contribute in the Baum-Welch fitting procedure.

Details

The Hidden Markov model uses a binomial test for the emission densities. Transition probabilities are modeled with a distance dependent decay, specified by the parameter transDist.

Value

A methimputeBinomialHMM object.

Examples

## Get some toy data
file <- system.file("data","arabidopsis_toydata.RData", package="methimpute")
data <- get(load(file))
print(data)
model <- callMethylation(data)
print(model)

methimpute documentation built on Nov. 8, 2020, 5:47 p.m.