Drug-Drug Interactions Predictions Using BiLSTM

Abstract

Various methods are innovated in the field of bioinformatics and drug discovery. The recent papers on drug discovery [23,25], insisted many ideas to the researchers of that field. In the field of drug discovery, ANN plays a vital role to differentiate drug discovery, Machine Learning and Deep Learning. These methods were used to predict interactions between D-D, T-T, D-T means Drug-Drug, Target-Target, Drug-Target interactions. In Machine Learning prediction is made by use of the Multilayer Perceptron, Decision Trees, Random Forest, Support Vector Machines and Naïve Bayes methods. In the other hand, Deep Learning make use of LSTM (Long Short Term Memory), GRUs and Transformers. Among all the methods, we are planned to implement BiLSTM with RDkit to predict the interactions between any two drugs.

Introduction

The paper reviews various neural network approaches, especially graph neural networks (GNNs), for drug-drug interaction (DDI) prediction and related tasks, citing models such as DDI Net, graph convolutional networks, MPNN-based GAT, and substructure-aware GNNs. Several studies focus on improving accuracy and handling challenges like unselected negative samples and prediction in inductive settings. Methods like BiLSTM and Support Vector Machines are also discussed for side effect and interaction predictions.

Materials and Methods:
The dataset is sourced from DrugBank, focusing on two selected drugs using their SMILES notation and unique IDs. The Blood Brain Barrier Penetration (BBBP) dataset is used for experiments. Morgan fingerprints generated from SMILES strings form the input features. A deep learning model is built with nine dense layers, 1D convolutional layers, max pooling, dropout, and a BiLSTM layer. Activation functions include ReLU for hidden layers and Softmax for the output layer. The model is trained using Adam optimizer and categorical cross-entropy loss with 10-fold classification.

Performance Evaluation:
Metrics such as accuracy (0.374), F1 score (0.32), AUC (0.865), average sensitivity (0.38), and specificity (0.93) are reported. Confusion matrix and ROC curves visualize the model's performance, showing strengths in specificity but lower accuracy and sensitivity.

Conclusion

A lot of different varieties of topics are discussed in the paper mentioned above. In our study, performance metrics is measured accurately with most effective scores. In future work, Graph Attention Network (GAT) and transformers might be used by the future researchers.

References

[1] Wei Di, Anurag Bhardwaj, Jianing Wel ., Deep Learning Essentials . [2] Chitta Rangan, Understanding Deep Learning [3] William L.Hamilton, Graph Representation Learning. [4] Josh Patterson and Adam Gibson by Deep Learning, A Practitioner’s Approach. [5] Charu C.Agarwal, Neural Networks and Deep Learning – A Text Book. [6] Michael Nielsen, Neural Networks and Deep Learning. [7] Lingfei wu, PengCui, Jian Pei, Liang Zhao, Graph Neural Networks: Foundations, Frontiers and Applications. [8] Ian Goodfellow, Aaron Courville, Yoshua Bengio, Deep Learning. [9] John D Kelleher, Deep Learning. [10] Anindya Halder, Biswanath Saha, Moumita Roy and Sukanta Majumder, A Novel Deep Sequential Learning architecture for drug-drug interaction prediction using DDINet,2025,1-15. [11] Konda Mani Saravanan, Jiang-Fan Wan, Liujun Zhang, John Z.H. Zhang, A Deep Learning based multi model approach for prediciting drug-like chemical compound’s toxicity,2024,164-175. [12] Yan Xia,Zilong Zhang, Quan Zou,Feifei cui, A Comprehensive review of deep learning based approaches for drug-drug interaction prediction,2025, 1-11. [13] Ramanathan Rajagopalan, Shalika M, Arubprasath M,Senthamarai R,A role of Artificial Intelligences in Drug Discovery and drug development- A critical review,2024, 1714-1723. [14] LiangchengDong, Baoming Feng, Zenqian Deng, Jin longWang, Peihao Ni, Yuanyuan Zhang, A Substructure-aware graph neural network incorporating relation features for drug-drug interaction prediction,2024,225-270. [15] Manel Gil-sorribes, Alexis Molina, Addressing Model Over complexity in drug drug interaction prediction with molecular fingerprints, 2025, 1-13. [16] Pietro Bongini, Elisa Messori, Niccolo Pancino,Monica Bianchini, A Deep learning approach to the prediction of drug-sideeffects on Molecular graphs,2022, 1-11. [17] Aga Basit Iqbal, Idris Afzal Shah, Injila, Assif Assad, Mushtaq Ahmed, Syed Zubair Shah, A review of deep learning algorithms for modelling drug interactions,2024,1-32. [18] Chiranjib Chakrabothy, Manojit Bhattacharya,Sang-Soo Lee, Zhi Hong Wen,Yi-Hao Lo, The changing scenario of drug discovery using AI to deep learning: Recent advancement, success stories, collaborations, nd challenges,2024,1-21. [19] Jing Zhu, Chao Che, Hao Jiang, Jian Xu, Jiajun Yin, Zhaoqian Zhong, SSF-DDI – a deep learning method utilizing drug sequence and substructure features for drug drug interaction prediction, 2024, 1-18 [20] Sabir Ali, Waleed Alam, Hilal Tyara, Kil to Chong, An accurate prediction of drug- drug interactions and sideeffects by using integrated convolutional and Bi LSTM networks,2025,1-9. [21] Monica Bianchini by Graph Neural Networks for the prediction of protein-protein interfsces,2020,127-132. [22] Yue-hua feng, Shao-Wu Zhang, Jian-Yu shi, DPDDI: a deep predictor for drug-drug interactions,2020, 1-15. [23] Sera Park, Comprehensive survey of Recent drug discovery using deep learning, 2021,1-36. [24] Arnold K.Nyamabo, Drug drug interaction prediction with learnable size-adaptive molecular substructures, 2022,1-12. [25] Fan Liang, Survey of Graph Neural Networks and applications, 2022,1-18.

Copyright

Copyright © 2025 Ms. K. Hema. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.