Abstract
Background: Defining specific molecular targets for cancer therapeutics remains a significant challenge in oncology. Many Food and Drug Administration (FDA)-approved anticancer drugs have incomplete target profiles, which limits our understanding of their mechanisms of action and opportunities for drug application. In this context, this study aimed to establish novel, biologically meaningful relationships between anticancer drugs and protein-coding genes. Methods: We developed a pharmacogenomic method that integrates transcriptomic data with drug activity data from the NCI-60 cancer cell line panel to study the interactions between 124 FDA-approved anticancer drugs and 399 cancer-related genes. Through this analysis, we identified gene-drug relationships and created a bipartite interaction network. To evaluate drug similarity, we developed a new index called the B-index. This novel similarity coefficient measures the association between two drugs based on their shared gene targets in the network. The index calculates the intersection of two sets of drug targets while considering the relative proportion of targets exhibited by each drug. For an independent assessment, we compared this network-based similarity with the chemical structural similarity of the drugs, computed based on two structural coefficients: Maximum Common Substructure and Tanimoto. Results: The study identified 1304 statistically significant drug-gene relationships, providing a large-scale network of pharmacogenomic interactions. Clustering analysis of the network, based on the B-index, grouped drugs with common targets together. This grouping was consistent with well-established drug classes and structural characteristics. Well-established drug pairs, such as cytarabine-gemcitabine or afatinib-neratinib, exhibited high B-index and structural similarity values, validating the methodology. Several novel gene associations were discovered, yielding testable hypotheses for mechanism-based repurposing. Conclusions: This work presents a comprehensive, network-based strategy for elucidating cancer drug targets by combining gene expression and drug activity profiles. Additionally, the B-index provides an alternative to conventional chemical similarity metrics, which can facilitate the identification of new therapeutic relationships and inform new drug applications and repositioning. These findings pave the way for the proposal of novel oncology drug targets.