An Optimised Mobilenet V2 Attention Parallel Network for Predicting Drug–Drug Interactions Through Combining Local and Global Features

Abstract

Drug–drug interactions (DDIs) are an important concern in the clinical practice and drug development process as these may lead to serious adverse effects on patient safety. Thorough DDI prediction is important for effective medication management and reduced risk factors. This work presents a new technique, namely MV2SAPCNNO: MobileNetV2 with simplicial attention network-based parallel convolutional neural network and narwhal optimiser, for improving the precision of DDI prediction. The proposed method starts with data preprocessing, including normalisation and noise reduction, to enhance the quality of the data. Then, MobileNetV2 with simplicial attention network (MV2SAN) is used to extract both local and global features from the dataset. These features are processed using a parallel convolutional neural network (PCNN), optimised by the narwhal optimiser (NO) to improve parameter tuning, minimise error and reduce computational complexity. The performance of the model is evaluated using accuracy, precision, recall and F-score. Experimental results prove that MV2SAPCN-NO achieves better performance over the current models of DDI prediction in accuracy and enhanced classification metrics. The narwhal optimiser enhances the model's convergence efficiency and decreases computational time with an excellent predictive performance. An efficient and accurate DDI prediction model was proposed called MV2SAPCNNO. This model actually outperformed traditional models, and such findings were exhibited to contribute towards secure medication administration, drug development processes and protection of patients in clinical practice.