Ronan Jasper G. Reponte, Joshua C. Rodriguez, Gerard D. Ompad, Angie M. Ceniza-Canillo
Drug discovery is a long and expensive process. Unfortunately, the pharmaceutical industry is currently facing a significant challenge of high attrition rates during drug development because most candidate molecules fail to pass toxicity studies, leading to limited availability of drugs in the market. The method of drug repurposing can be employed to address this issue, which recognizes new applications of existing drugs outside of their original intended purpose. This paper utilizes a novel complex multi-layered network (MLN) approach for aggregating drug networks of varying information into a single unified network. Under the hypothesis that drugs that are closely related will appear as neighboring nodes, using a community detection approach allows novel discovery of new applications of old and existing drugs. This paper describes a method to aggregate MLNs. Drug data is collated from DrugBank and PubChem; an adjacency matrix is created, from which the Normalized Graph Laplacian (NGL) is generated. Prior to network aggregation, eigenvectors, and Uniform Manifold Approximation and Projection (UMAP) methods were utilized to reduce the network dimension. Additionally, a nonparametric inference was implemented to test for complete spatial randomness (CSR) to check for node scattering. This novel method showed some interesting insights into old and existing drugs with respect to their neighboring drugs by revealing potential repurposing opportunities based on their proximity within the network, suggesting that closely grouped drugs may share therapeutic effects, or other properties. © 2024 IEEE.
University of San Carlos, Department of Computer, Information Sciences and Mathematics, Cebu City, Philippines
