A tandem white shark algorithm approach for optimizing drug–disease and drug–drug interactions in multimorbidity and polypharmacy

Abstract

Minimizing Drug–disease and drug–drug Interactions in Polypharmacy and Multimorbidity (DIPM) is a challenging problem. The challenge is caused by the large number of similar medications that can be prescribed for each disease and the adverse reactions that may be experienced by the selected medication(s) for one or more of the diseases and medications identified. Therefore, it is a selection problem as to which medication should be prescribed to the patient for each disease while maintaining the least drug–disease and drug–drug interactions. This is an optimization problem, and to the best of our knowledge, it has never been addressed before using search-based algorithms. The aims of this study are to explore and find the best performance among the employed algorithms in providing optimal or near-optimal solutions to this problem, and then to fine-tune an optimization approach to DIPM. Four algorithms are consequently employed, which are: Dragonfly Algorithm (DA), Whale Optimization Algorithm (WOA), Genetic Algorithm (GA), and White Shark Optimization (WSO). To test the most suitable one among the algorithms, two synthetic datasets are used, representing simple and more complex problems. The results, considering only drug–disease interactions, show that GA is outperforming the others for a simple problem. However, WSO is found to outperform the others when the problem is more complex. The WSO is accordingly formed into a tandem approach, considering both drug–disease and drug–drug interactions for a real-world problem dataset. The approach indicates promising results with accuracy of (100%) for drug–drug interaction and (96.6%) for drug–disease interaction in (5.32) seconds.