Fig 5. Predictive importance of training features across four common cell lines.

The method permuting out-of-bag predictor observation was adopted to evaluate the relative importance of each feature. All features showed good robustness with effective importance (> = 0.05) across all cell lines. Some features such as DIS contribute > = 38% in all cells. However, some features such as WEEC and EWS have different importance in different cells.