Coefficients for final gene expression models

Coefficients for elastic net models fit using various -omics data types on the TCGA dataset, for the 217 target genes in the "all data types" analysis. 

Columns correspond to target genes (gene symbols), and rows correspond either to principal components (for 27K and 450K methylation), -omics features (all other data types), or covariates (cancer type indicator variables or log(mutation burden)).

An 'NA' value in a cell indicates that feature was not used in the model for the corresponding gene (for a gene feature this means that gene was not in the top 8000 genes by MAD, for a cancer type feature this means that cancer type was not included in the training set based on our mutation filters). A 0 value in a cell indicates that feature was included in model training, but it was not selected by the elastic net feature selection algorithm.

Hyperparameter selection procedure is described in the methods section of the paper: https://greenelab.github.io/mpmp-manuscript/