Community Detection on a Modified Adjacency Matrix: A Novel Network Approach in Drug-Drug Interaction Publisher



Shamami MA¹ ; Teimourpour B² ; Sharifi F³ Show Affiliations
Source: 2024 20th CSI International Symposium on Artificial Intelligence and Signal Processing# AISP 2024 Published:2024

Abstract

In pharmacotherapy and drug discovery, understanding potential interactions between drugs is crucial for providing patient safety and optimizing treatment outcomes. While much research has focused on identifying drug-drug interactions (DDIs), there is a growing need to uncover non-interacting drug pairs. Non-interactions can be as significant as interactions, offering valuable insights into safe co-prescription practices. In this study, we introduce a groundbreaking approach to DDI detection by employing community detection methods: a clustering approach applied to a modified adjacency matrix. We then used the silhouette criteria to determine the optimal number of clusters, ensuring that non-interacting drugs are accurately represented in distinct groups. This method leverages the inherent structure of the modified adjacency matrix to differentiate between interacting and non-interacting drug pairs accurately. We demonstrated the effectiveness of our method through extensive experiments on real-world datasets, achieving an accuracy rate of 95.3% in predicting non-interactions. Our results highlight our proposed technique's reliability and efficiency in identifying safe drug combinations while minimizing the adverse effects of unintended interactions. © 2024 IEEE.