deconvolute {granulator}R Documentation

Deconvolution from bulk RNAseq

Description

deconvolute predicts cell type proportions from bulk RNAseq data by applying multiple deconvolution methods.

Usage

deconvolute(m, sigMatrix, methods = get_decon_methods(), use_cores = 1)

Arguments

m

Bulk RNAseq: a genes (rows) by samples (columns) matrix containing transcript-per-million (TPM)-normalized gene expression values.

sigMatrix

Reference profile: a matrix or a named list of matrices. Each signature matrix should be a genes (rows) by cell types (columns) data frame containing TPM-normalized gene expression values of signature genes.

methods

Deconvolution methods: a character vector containing the names of the deconvolution methods to be applied. By default, all methods are run. Functions are either reimplementations of published methods or wrapper functions for published packages:

  • ols: ordinary least squares

  • nnls: non negative least squares regression model. Adapted from Abas et al. (2009)

  • qprog: quadratic programming without constraints

  • qprogwc: quadratic programming non-negative and sum-to-one constraints. Adapted from Gong et al. (2015)

  • dtangle: wrapper for the cell deconvolution function dtangle form the package dtangle

  • rls: robust linear regression. Adapted from Monaco et al. (2019)

  • svr: support vector regression. Adapted from Newman et al. (2015)

use_cores

Number of cores to use for parallel processing

Value

Returns a list containing two elements:

Author(s)

Vincent Kuettel, Sabina Pfister

Examples

# load demo PBMCS data
load_ABIS()

# generate list of reference profiles to be tested
sigMatrix <- list(
sig1 = sigMatrix_ABIS_S0, 
sig2 = sigMatrix_ABIS_S1)

# deconvolute
decon <- deconvolute(m = bulkRNAseq_ABIS, 
sigMatrix = sigMatrix)


[Package granulator version 1.0.0 Index]