Figure 4.
Deep learning analysis of intron groups. (A) Flow diagram for training the deep neural network. (B) Performance of the model in distinguishing introns of different groups. ROC curves were plotted for individual pairwise comparisons with AUC values shown in parentheses. (C) t-SNE plots of the 15 genomic features most predictive for distinguishing intron groups. Features distinguishing Group A from Groups C and D are shown above and those distinguishing Group C from Group D below. Features colored blue or red indicate the group for which they are positively correlated.
