Fig. 1
From: DeepM6ASeq: prediction and characterization of m6A-containing sequences using deep learning
A graphic illustration of DeepM6ASeq model structure. The genome sequence (A in red represents an m6A site) is first one-hot encoded as input, then the input is sequentially fed into two layers of CNN in order. The first CNN layer functions as a motif detector while the second CNN layer captures features of a higher level. After the CNN layers is one BLSTM layer to capture sequential order. The output units of the BLSTM layer are followed by the fully connected layer, and finally the model outputs the prediction result