Fig. 6.

Exploration of LR feature weights. Heatmap of feature weights learned by the LR model in two different cell types (ESC-H1 and HepG2) using three different feature selection methods (ElasticNet, L1, and RFE) to obtain ~ top 150 features each. Rows are clustered using hierarchical clustering with uncentered Pearson correlation as the similarity measure and average linkage. Pullouts show four groups of features with the indicated characteristics