Abstract:Detecting the unintended adverse reactions of drugs (ADRs) is a crucial concern in pharmacological research. The experimental validation of drug–ADR associations often entails expensive and time-consuming investigations. Thus, a computational model to predict ADRs from known associations is essential for enhanced efficiency and cost-effectiveness. Here, we propose BiMPADR, a novel model that integrates drug gene expression into adverse reaction features using a message passing neural network on a bipartite graph of drugs and adverse reactions, leveraging publicly available data. By combining the computed adverse reaction features with the structural fingerprints of drugs, we predict the association between drugs and adverse reactions. Our models obtained high AUC (area under the receiver operating characteristic curve) values ranging from 0.861 to 0.907 in an external drug validation dataset under differential experiment conditions. The case study on multiple BET inhibitors also demonstrated the high accuracy of our predictions, and our model's exploration of potential adverse reactions for HWD-870 has contributed to its research and development for market approval. In summary, our method would provide a promising tool for ADR prediction and drug safety assessment in drug discovery and development.

What problem does this paper attempt to address?

The paper attempts to address the problem of predicting adverse drug reactions (ADRs) for new drugs. Specifically, the authors propose a deep learning framework named BiMPADR, which aims to predict the association between new drugs and adverse reactions by integrating drug gene expression and structural fingerprint information. This method aims to improve the efficiency and cost-effectiveness of predictions, reducing the time and resource consumption of experimental validation. ### Background - **Importance of Adverse Reactions**: Adverse drug reactions (ADRs) are a major challenge in drug development, leading not only to illness and death but also increasing the economic burden on public health systems. Statistics show that nearly 100,000 people die each year due to adverse drug reactions, with the incremental total cost per patient ranging from 702 to 7318 euros. - **Limitations of Existing Methods**: Traditional prediction methods often rely on pharmacological experiments or clinical observations, which are time-consuming and resource-intensive, and often fail to detect adverse reactions of new drugs. - **Development of Computational Methods**: In recent years, computational prediction methods, especially deep learning techniques, have made significant progress in predicting adverse reactions using drug-related databases. ### Research Objectives - **Improve Prediction Accuracy**: Enhance the accuracy of predicting adverse reactions of new drugs by integrating drug gene expression and structural fingerprint information. - **Overcome Limitations of Existing Methods**: Existing methods have limitations in handling new drugs due to the lack of sufficient prior information. The BiMPADR framework aims to address this issue by providing a predictive framework suitable for new drug molecules. ### Methods - **Data Sources**: Utilize publicly available data, including drug gene expression data and structural fingerprint data. - **Model Architecture**: Employ a bipartite graph-based message passing neural network (MPNN) to combine drug gene expression features with structural fingerprint features to predict the association between drugs and adverse reactions. - **Performance Evaluation**: Evaluate the model using multiple datasets, including training sets, test sets, and external validation datasets, assessing the model's AUC value, precision, and accuracy. ### Results - **High Prediction Accuracy**: The model achieved AUC values ranging from 0.861 to 0.907 on external validation datasets, indicating high prediction accuracy. - **Stability**: The model's performance is stable under different input conditions, with minimal impact from different types of compound fingerprints and cell line expression data. - **Comparative Experiments**: Compared to existing methods (such as Pauwels' SCCA and DrugClust), BiMPADR performs well in terms of AUC value, although it is slightly lower than SCCA in some metrics, its overall performance is superior. ### Conclusion - **Potential Applications**: The BiMPADR framework provides a powerful tool for predicting adverse reactions of new drugs, aiding in the early assessment of potential drug safety, and reducing drug development costs and health risks. - **Future Research Directions**: Further explore the role of gene expression information in predictions, and optimize the robustness and generalization ability of the model.

BiMPADR: A Deep Learning Framework for Predicting Adverse Drug Reactions in New Drugs

Predicting adverse drug reactions through interpretable deep learning framework

Prediction of Adverse Drug Reactions by Combining Biomedical Tripartite Network and Graph Representation Model.

Predicting Drugs Suspected of Causing Adverse Drug Reactions Using Graph Features and Attention Mechanisms

ADRNet: A Generalized Collaborative Filtering Framework Combining Clinical and Non-Clinical Data for Adverse Drug Reaction Prediction

Advancing Adverse Drug Reaction Prediction with Deep Chemical Language Model for Drug Safety Evaluation

Research on Adverse Drug Reaction Prediction Model Combining Knowledge Graph Embedding and Deep Learning

Prediction of adverse drug reactions based on knowledge graph embedding

Abstract 4923: A systematic framework for predicting adverse drug reaction signals using medical data

Developing A Deep Learning Natural Language Processing Algorithm For Automated Reporting Of Adverse Drug Reactions

Deep Learning Prediction of Adverse Drug Reactions Using Open TG-GATEs and FAERS Databases

A Survey on Computational Approaches to Predicting Adverse Drug Reactions.

Identifying the serious clinical outcomes of adverse reactions to drugs by a multi-task deep learning framework

Establishment of a Rapid Identification of Adverse Drug Reaction Program in R Language Implementation Based on Monitoring Data

Prediction of adverse drug reactions due to genetic predisposition using deep neural networks

Predicting Adverse Drug Reactions from Social Media Posts: Data Balance, Feature Selection and Deep Learning

ADEP: A Novel Approach Based on Discriminator-Enhanced Encoder-Decoder Architecture for Accurate Prediction of Adverse Effects in Polypharmacy

Mining Real-World Big Data to Characterize Adverse Drug Reaction Quantitatively: Mixed Methods Study

In Silico Search Of Putative Adverse Drug Reaction Related Proteins As A Potential Tool For Facilitating Drug Adverse Effect Prediction

mtADENet: A novel interpretable method integrating multiple types of network-based inference approaches for prediction of adverse drug events

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