Fig. 1: The DARTS computational framework. | Nature Methods

Fig. 1: The DARTS computational framework.

From: Deep-learning augmented RNA-seq analysis of transcript splicing

Fig. 1

a, Overall workflow of DARTS. b, Schematic of the DARTS DNN features, including cis sequence features and trans RBP features. c, Overview of training and leave-out RBPs, and the number of significant differential splicing events called by DARTS BHT(flat) on the ENCODE data (illustrated by bar charts above the outer and middle circles). We used 196 RBPs knocked down in both the K562 and HepG2 cell lines for training (orange), while the remaining 58 RBPs knocked down in only one cell line were leave-out data (light orange) (illustrated in the inner circle). RRM, RNA recognition motif; KH, K homology; ZNF, zinc finger. d, Comparison of the DARTS DNN with baseline methods in leave-out datasets. KD, knockdown; CTRL, control; RPL23A, ribosomal protein L23a; AQR, aquarius intron-binding spliceosomal factor.

Back to article page