Abstract:Drug-drug interactions (DDIs) and their associated adverse drug reactions (ADRs) represent a significant detriment to the public health. Existing research on DDIs is primarily focused on pairwise DDI detection and prediction. It is highly needed to develop effective computational tools for high-order DDI prediction. Here we show that deep learning can be effectively utilized to predict ADRs induced from high-order DDIs. In this manuscript, we present a deep learning model D3I for cardinality-invariant and order-invariant high-order DDI prediction. The D3I models achieve 0.740 F1 value and 0.847 AUC value on balanced high-order DDI prediction, and outperform other models on order-2 DDI prediction. These results demonstrate the strong potential of D3I and deep learning models in tackling the prediction problems of high-order DDIs and their induced ADRs. In addition, D3I is able to derive single drug representations, which conform to our current knowledge on single drugs, from their behaviors in drug combinations. D3I can also correctly predict ADRs for drug combinations in which no single drugs on their own induce ADRs, and improve ADR prediction on drug pairs by learning from all drug combinations. To the best of our knowledge, D3I is the first deep model for high-order DDI prediction.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve a key problem in the prediction of drug - drug interactions (DDIs) and their associated adverse drug reactions (ADRs), especially the prediction of high - order DDIs (interactions involving more than two drugs). Specifically: 1. **Limitations of existing research**: Most of the existing DDI research mainly focuses on the detection and prediction of pairwise DDIs (i.e., second - order DDIs), while the research on high - order DDIs involving multiple drugs is very limited. However, in actual clinical applications, high - order DDIs account for a considerable proportion, especially among the elderly, many of whom take two or more drugs daily. 2. **Challenges in high - order DDI prediction**: - **Complexity**: The prediction of high - order DDIs is more complex than that of low - order DDIs because only a part of the drugs in the combination may cause ADRs, and these drugs may be safe when used alone. - **Data sparsity**: There is relatively little data on high - order DDIs, and it is difficult to obtain sufficient training samples. - **Model invariance**: It is necessary to develop a model that can handle any number of drug combinations and is not affected by the input order of drugs. 3. **Solution**: The authors propose a new deep - learning model D3I (DeepDDI), which can predict ADRs caused by high - order DDIs without considering the cardinality of drug combinations (i.e., the number of drugs in the combination) and the input order. D3I achieves this goal in the following ways: - **Encoder**: Each drug is encoded into a latent representation (embedding). - **Aggregator**: A single high - level representation of the drug combination is learned from the embeddings of individual drugs. - **Predictor**: The representation of the drug combination is used to predict the likelihood of ADRs. 4. **Experimental results**: D3I has been extensively experimented on two public high - order DDI datasets, achieving an F1 value of 0.740 and an AUC value of 0.847, and also outperforming other models in second - order DDI prediction. In addition, D3I can generate single - drug representations that are consistent with current drug knowledge and improve the ADR prediction of drug pairs by learning all drug combinations. In summary, the main objective of this paper is to develop an effective computational tool to address the important but not yet fully studied problem of high - order DDI and the prediction of ADRs caused by them.

Deep Learning for High-Order Drug-Drug Interaction Prediction

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

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

DPDDI: a deep predictor for drug-drug interactions

A Novel Deep Learning Model for Drug-drug Interactions

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

TP-DDI: A Two-Pathway Deep Neural Network for Drug–Drug Interaction Prediction

MLRDA: A Multi-Task Semi-Supervised Learning Framework for Drug-Drug Interaction Prediction

3DGT-DDI: 3D graph and text based neural network for drug-drug interaction prediction

Geometric molecular graph representation learning model for drug-drug interactions prediction

MASMDDI: multi-layer adaptive soft-mask graph neural network for drug-drug interaction prediction

Enhancing drug–drug interaction prediction by three-way decision and knowledge graph embedding

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

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

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

CNN-DDI: a learning-based method for predicting drug–drug interactions using convolution neural networks

Drug-drug Interaction Prediction with Graph Representation Learning

Drug-Drug Interaction Predicting by Neural Network Using Integrated Similarity

Prediction of multiple types of drug interactions based on multi-scale fusion and dual-view fusion

Drug-drug interaction prediction based on co-medication patterns and graph matching

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