EMeth

This package provides implementations of the EMeth algorithm. It contains two functions.

emeth, which provides implementation of two families of EMeth, normal and laplace, accordingto what likelihood is used. Generally we recommend the use of laplace family.
cv.emeth, which helps tuning the ridge penalty by cross validation.

Prerequisite

EMeth package requires R version 3.6.0 or higher. It requires the package quadprog

Installation

To install this package in R, use

    library("devtools");
    install_github("Hanyuz1996/EMeth")

Example

An example use:

     library("EMeth")
     em1 = cv.emeth(Y, eta, mu, aber = TRUE, V='c', init = 'default',
        family = 'laplace', nu = penalty, folds = 5, maxiter = 50, verbose = TRUE)
     cell_type_prop_est = em1$result$rho

where

Y is the DNA methylation data matrix with rows for CpG probes and columns for samples,
eta is a vector of tumor purity that can be set as 0 for non-tumor studies,
mu is the reference data for cell type-specific DNA methylation, with each column for one cell type, and
nu is the penalty parameters, for example. we use nrow(Y)*(10^seq(-2,1,1)) in our TCGA analysis.

This function cv.emeth automatically runs the cross validation procedure. Please see the help document for this function ?cv.emeth for more details of these and other parameters. Example data are provided in the example folder.

The output is a list with three elements

result: The result of the EMeth algorithm using the penalty value selected by cross-validation. It is a list and documentation of its entries can be found in the help file for function emeth. The following entries of result may be of interest.
- rho: a matrix of the cell type proportion estimates (rows for samples and columns for cell type).
- gamma: a matrix whose (i,j)-th entry is the probability that the i-th probe in the j-th sample is aberrant (i.e., the DNA methylation is the j-th bulk sample is not consistent with the deconvolution model and the cell type-specific methylation reference). This matrix could be used by other methods to select the set of CpG probes to be used for deconvolution.
- nu0: estimates of DNA methylation in the special cell type without reference, i.e., tumor cells in bulk tumor samples that include tumor cells as well as other cell types such as tumor-infiltrating immune cells.
choosenu: The value of the nu (the penalty) chosen by the cross-validation.
losslist: A matrix saving the loss for each fold and each choice of nu.

Support

You can contact zhanghy1996@gmail.com

Citation

If you use the software, please cite our paper: Zhang et al. 2021. The pipelines for simulation studies and real data analysis in this paper are contained in this repository.

Authors

Hanyu Zhang (University of Washington)

Wei Sun (Fred Hutchinson Cancer Research Center)

Reference

Zhang et al. 2021, EMeth: An EM algorithm for cell type decomposition based on DNA methylation data.

Hanyuz1996 / EMeth

readme