Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: Predicting protein-ligand binding residues with deep convolutional neural networks

Fig. 2

Architecture of the deep convolutional neural network in std-DeepCSeqSite (stdDCS-SI). Each cell represents a dimension of a representation. The m×d representation of an amino acid sequence is the input of the network, where m is the length of the amino acid sequence, and d is the dimension number of the feature space. Block (k×1,2c) represents a BasicBlock with a k×1 kernel size and 2c output channels, and the structure of Plain (k×1,2c) is the same as that of Block (k×1,2c) without residual connection. The situation of k=3, stride =1 and c=3 is described in this figure. Each m×1 cell grid represents the output of a convolution kernel. The right-most representation is the input for the softmax

Back to article page