Fig. 4
From: Dissecting splicing decisions and cell-to-cell variability with designed sequence libraries
Accurate prediction of splicing ratios. R2 scores for predicting splicing ratios of a set of variants not used at any point to build the model (10% of library variants) by Gradient Boosting Regression with the indicated sets of features (top) and measured splicing ratios vs. predictions based on a combination of hexamer counts, RBP motif scores, MaxEntScan scores and secondary structure features (bottom) for the indicated splicing types
