R script implementing regression analysis to estimate the proportion of cells from mixture gene expression data. Analysis is based on reference expression profiles and linear regression with Elastic net regularization is used.
Function for running the regression analysis is regressionAnalysis. Input for the function is normalized gene expresison matrix of mixture for which the relative cell proportions are estimated and normalized gene expression matrix of reference cell types.
proportions <- regressionAnalysis(mixture = myMixture, reference_cells = myReference)The value of Elastic-net mixing parameter alpha can be changed. Default value for alpha is 0.5.
proportions <- regressionAnalysis(mixture = myMixture, reference_cells = myReference, alpha = 0.1)proportions contains the estimated relative cell proportions for each mixture sample present in the matrix myMixture.
Directory data contains normalized example data (Abbas et al. PLoS One. 4:e6098, 2009). The file GSE11058_mixtures.txt contains the cell line mixtures and the file GSE11058_reference_cells.txt contains reference cell types (Jurkat, IM9, Raji, THP1, median sample).
Suvi Luoto (suvi.luoto@uta.fi).