Skip to main content
. 2019 Jul 5;35(14):i269–i277. doi: 10.1093/bioinformatics/btz339

Fig. 1.

Fig. 1.

Overview of the deep learning architectures evaluated in this work. These include CNN-only models known for their ability to detect motifs (left), RNN-only models (center) which excel at capturing long-term sequence dependencies, and hybrid CNN-RNN models. The input for all variants is either a one-hot encoding or a k-mer embedding of the DNA/RNA sequence obtained using word2vec