What problem does this paper attempt to address?

The problem that this paper attempts to solve is to predict drug - drug interactions (DDI) in drug combination therapies. Specifically, existing methods often ignore the sequence information of drug molecules or fail to comprehensively consider the influence of molecular sub - structures on DDI when predicting potential DDI. This has led to limitations in the prediction accuracy of existing models, especially when dealing with unknown drugs. To solve these problems, the authors propose a new deep - learning - based model, SSF - DDI (Sequence and Substructure Features for Drug - Drug Interaction prediction), which combines the sequence features and sub - structure features of drug molecules to improve the accuracy and comprehensiveness of DDI prediction. ### Main contributions 1. **Proposed a new DDI prediction model, SSF - DDI**: This model integrates the sequence features and structural features of drug molecules. By introducing the topological properties and sequence features of drug molecules, it captures a wider range of feature information, thereby achieving a more accurate and comprehensive representation of drug molecule features. 2. **Introduced a new drug sub - structure graph feature encoder (SGFE)**: This encoder can effectively extract drug atom and molecular structure features. 3. **Conducted experiments in the transductive setting and the inductive setting**: The experimental results show that SSF - DDI significantly outperforms other methods on multiple real - world datasets. Especially in the inductive setting, SSF - DDI performs particularly well in predicting the DDI of new drugs, with an accuracy improvement of 5.67%. ### Method overview - **Drug molecule sequence feature extraction module**: Use a convolutional neural network (CNN) to extract the sequence features of drug molecules, and determine the interaction score between the two drug sequence features through a mixed attention mechanism (MixAttention). - **Sub - structure feature extraction module**: Construct a new drug sub - structure graph feature encoder (SGFE), use a directed message - passing neural network (D - MPNN) to extract sub - structure features, and then generate a feature vector containing sub - structure and topological information through a multi - layer graph attention network (GAT) and self - attention graph pooling (SAGPooling). - **Prediction module**: Fuse the extracted sequence features and sub - structure features and input them into a fully - connected layer for final drug relationship prediction. ### Experimental results - **Experiments were conducted on the DrugBank and Twosides datasets**: The results show that the accuracy of SSF - DDI in the transductive setting reaches 96.45%, with a relative accuracy improvement of 0.36%; in the inductive setting, the accuracy of SSF - DDI in predicting the DDI of new drugs reaches 87.3%, with an accuracy improvement of 5.67%. - **Evaluation metrics**: Six evaluation metrics were used, including accuracy (ACC), area under the ROC curve (AUC), F1 - value (F1), precision (Precision), recall (Recall), and average precision (AP). ### Conclusion SSF - DDI significantly improves the accuracy and robustness of DDI prediction by combining the sequence features and sub - structure features of drug molecules, especially when dealing with unknown drugs. This method has important significance in practical applications and is helpful for guiding drug development and clinical drug use safety.

SSF-DDI: a deep learning method utilizing drug sequence and substructure features for drug–drug interaction prediction

SSI–DDI: substructure–substructure interactions for drug–drug interaction prediction

Deep learning for drug‐drug interaction prediction: A comprehensive review

STNN-DDI: A Substructure-aware Tensor Neural Network to Predict Drug-Drug Interactions

Comprehensive evaluation of deep and graph learning on drug–drug interactions prediction

Molecular Structure-Based Double-Central Drug-Drug Interaction Prediction

DSN-DDI: an accurate and generalized framework for drug–drug interaction prediction by dual-view representation learning

An effective framework for predicting drug–drug interactions based on molecular substructures and knowledge graph neural network

Comprehensive evaluation of deep and graph learning on drug-drug interactions prediction

DeepDrug: A General Graph-Based Deep Learning Framework for Drug-Drug Interactions and Drug-Target Interactions Prediction

Learning size-adaptive molecular substructures for explainable drug–drug interaction prediction by substructure-aware graph neural network

Multi-View Substructure Learning for Drug-Drug Interaction Prediction

DPDDI: a deep predictor for drug-drug interactions

Multitype drug interaction prediction based on the deep fusion of drug features and topological relationships

Molecular Substructure-Aware Network for Drug-Drug Interaction Prediction

Deep learning improves prediction of drug–drug and drug–food interactions

StructNet-DDI: Molecular Structure Characterization-Based ResNet for Prediction of Drug–Drug Interactions

MSDAFL: molecular substructure-based dual attention feature learning framework for predicting drug-drug interactions

DSIL-DDI: A Domain-Invariant Substructure Interaction Learning for Generalizable Drug–Drug Interaction Prediction

A Molecular Fragment Representation Learning Framework for Drug–Drug Interaction Prediction

R2-DDI: relation-aware feature refinement for drug-drug interaction prediction