Integrative Cancer Pharmacogenomics to Infer Large-Scale Drug Taxonomy

Abstract

Identification of drug targets and mechanism of action (MoA) for new and uncharacterized anticancer drugs is important for optimization of treatment efficacy. Current MoA prediction largely relies on prior information including side effects, therapeutic indication, and chemoinformatics. Such information is not transferable or applicable for newly identified, previously uncharacterized small molecules. Therefore, a shift in the paradigm of MoA predictions is necessary toward development of unbiased approaches that can elucidate drug relationships and efficiently classify new compounds with basic input data. We propose here a new integrative computational pharmacogenomic approach, referred to as Drug Network Fusion (DNF), to infer scalable drug taxonomies that rely only on basic drug characteristics toward elucidating drug-drug relationships. DNF is the first framework to integrate drug structural information, high-throughput drug perturbation, and drug sensitivity profiles, enabling drug classification of new experimental compounds with minimal prior information. DNF taxonomy succeeded in identifying pertinent and novel drug-drug relationships, making it suitable for investigating experimental drugs with potential new targets or MoA. The scalability of DNF facilitated identification of key drug relationships across different drug categories, providing a flexible tool for potential clinical applications in precision medicine. Our results support DNF as a valuable resource to the cancer research community by providing new hypotheses on compound MoA and potential insights for drug repurposing. .

Integrative Cancer Pharmacogenomics to Establish Drug Mechanism of Action: Drug Repurposing

“…DNF is the first tool to integrate pharmaco-chemo-genomics high-throughput data, including chemical structures, pharmacological growth inhibition profiles and post-treatment transcriptional perturbations, to develop comprehensive drug taxonomy [20]. DNF leverages PharmacoGx, our R/Bioconductor package including the largest pharmacogenomic datasets published to date [21].…”

“…For the second data layer, DNF capitalizes on the recent L1000 dataset composed of 1.8 M gene expression profiles pre-and post-treatment for over 20,000 chemical compounds [15]. These transcriptional perturbation data are used to identify genes that are significantly down and upregulated due to drug treatment, referred to as drug perturbation signatures [20,21]. These signatures can in turn be used to find compounds with similar effect on the cell's transcriptome.…”

“…To validate the performance of DNF versus single layers, we tested the predictive value against gold standard benchmarks: DrugBank for drug-targets associations [27] and ChEMBL for anatomical therapeutic classification [28]. We found that DNF significantly outperformed all single layers and showed the highest accuracy in detecting drug pairs with the same MoA when using both drug-target and anatomical therapeutic classification benchmarks [20].…”

“…Moreover, these approaches integrate a limited number of data layers, despite their high complementarity [18,19]. This motivated us to develop Drug Network Fusion (DNF), a computational method combining diverse types of drug-centered data to improve drug taxonomy for inference of drug target and MoA [20].…”

Large-scale integration of heterogeneous pharmacogenomic data for identifying drug mechanism of action

“…However, the sheer amount and heterogeneity of these multi-omics data pose great challenges in the integration process: (i) the mixed formats, scales, and metrics, (ii) the complementary but high-dimensional information, and (iii) the incomplete and noisy nature of these datasets. As far as we know, Drug Network Fusion (DNF) 11 was the only previous attempt to simultaneously integrate the drug structure, perturbation and sensitivity data. Notably, DNF used a similarity network fusion approach, 12 in which a similarity network is constructed for each input data sources, and these similarity networks are then iteratively fused together until convergence to obtain a single similarity network.…”