Drug-Drug Interaction: An Improved Prediction Approach Based on Convolutional Neural Networks
K. Soni Sharmila, Thanga Revathi S, Kiran Sree Pokkuluri · 2023
Drug-Drug Interaction (DDI) is a significant challenge in modern healthcare as they have the potential to cause adverse side effects and hinder patient well-being. Accurate DDI prediction is critical for ensuring the efficacy and safety of medication management. This article proposes a novel method for detecting DDIs that makes use of convolutional neural networks (CNNs) for performing feature extraction and prediction augmentation. CNNs are used to automatically extract useful characteristics from drug combinations by using the underlying data patterns within drug interaction datasets. These collected features are then incorporated into a prediction model, which allows for more accurate detection of possible DDIs. To compare the performance of the proposed method, traditional classifiers used for drug-drug prediction, such as support vector machines (SVM), adaptive boosting (AB), and gradient boosting decision tree (GBDT), were preferred. The simulation results clearly demonstrated the efficacy of the proposed approach, emphasizing its potential for significantly enhancing the prediction accuracy. Notably, when multiple characteristics were interacted, the proposed strategy improved drug feature extraction by 24.4% when compared to the utilization of single features. This improvement indicates the method's robustness and capacity to capture complex interactions between pharmacological features, resulting in more accurate predictions.