emeth: emeth: cell type decomposition from DNA methylation data...

In Hanyuz1996/EMeth: Cell type decomposition based on methylation data

Description

This function is an implementation of EMeth algorithm. It provides estimation of cell type proportion for each sample in bulk tissue samples Y. It needs a reference matrix mu to conduct the estimation. EMeth has two families, normal and laplace, depending on what likelihood is used. It also has a penalization term (ridge penalty).

Usage

emeth(Y, eta, mu, aber, V, init = "default", family = "laplace", nu = 0, maxiter = 50, verbose = FALSE)

Arguments

Y	Matrix of size K*I. Methylation of bulk sample for which the cell type decomposition is to be estimated. Y is usually a K*I matrix where K is the number of probes used and I is the number os samples.
eta	Vector of size I. Tumor purity of each sample.
mu	Matrix of size K*Q. Reference matrix, provided by literature. A sample reference data in the example directory is provided, that is obtained from six previous literatures.
aber	Logic variable: if there is unknown aberrant cell type.
V	string: default to be 'c' which stands for constant weight for all probes. It might be 'b' for binomial variance structure or 'w' for specific weight structure of variance.
init	If init is a string 'default', we will adopt the default random initialization of all parameters for the algorithm. Otherwise one can provide a list of initialization of parameters.
family	string: accept 'normal' or 'laplace' to specify what likelihood will be used in the algorithm.
nu	a nonnegative number that stands for the penalty.
maxiter	max time of iteration of the EM algorithm, default to be 50.
verbose	logic variable. If TRUE then will print additional information in iteration of EMeth.

Value

rho	estimated proportion of cell types in each sample.
sigma_c	estimated variance.
lambda	estimated ratio between aberrant variance and consistent variance.
nu0	estimated methylation level of unknown cell type.
pi_a	estimates of the proportion of aberrant probes in each sample.
gamma	estimated probability of each probe in each sample to be aberrant.
weights	the working weights used in the iteration of EMeth.
iter	Number of iterations.

Author(s)

Hanyu Zhang

Examples

## See examples folder

Hanyuz1996/EMeth documentation built on Dec. 31, 2020, 12:59 p.m.