Fig. 1From: SSF-DDI: a deep learning method utilizing drug sequence and substructure features for drug–drug interaction predictionOverview of SSF-DDI method. SSF-DDI sequence feature encoder captures sequence features are extracted from drug molecules using multilayer convolutional neural networks (CNNs) and MixAttention. The substructure feature extraction module employs directed message-passing neural network (D-MPNN) to extract substructures. Then, the extracted substructure information is passed through multilayer graph attention network (GAT) and self-attention graph pooling (SAGPooling) layer to generate feature vectors containing both substructure and topology information. Subsequently, the prediction module predicts drug interactions based on these extracted featuresBack to article page