Membrane Protein Amphiphilic Helix Structure Prediction Based on Graph Convolution Network

The amphiphilic helix structure in membrane proteins is involved in membrane-related biological processes and has important research significance. In this paper, we constructed a new amphiphilic helix dataset containing 70 membrane proteins with a total of 18,458 amino acid residues. We extracted three commonly used protein features and predicted the membrane proteins amphiphilic helix structure using graph convolutional neural network. We improved the prediction accuracy of membrane proteins amphiphilic helix structure with the newly constructed dataset by rigorous 10-fold cross-validation.