Fig. 1
From: Deep convolutional neural networks for pan-specific peptide-MHC class I binding prediction
Schematic representation of overall training process of the DCNN. An ILA is converted from peptide binding information of training dataset. The DCNN extracts low-level features from the ILA and combines them into high-level features(motifs) through multiple convolutional and pooling layers. The DCNN learns these high-level features to be used for classifying the input ILA into binder or non-binder through fully connected layers