Figure 3

Model for prediction of clinical stage using regularized regression. (A) We used an integrative matrix of 5,188 genomic features from 1,122 genes in order to predict clinical stage using elastic net regularized regression. (B) The top 15 features ranked by coefficient in the model are shown. (C) Methylation features dominate the predictor set, but gene expression, copy number alteration and mutation features also play a critical role. The darker color indicates the direct association while the lighter color represents the inverse association.