Network biology approaches to understand patients’ response to cancer treatments

Abstract

Network-based approaches can be used to successfully detect patient responses to cancer treatment when integrated with multi-omics data. We have recently reported that protein-protein interaction (PPI) networks can be leveraged to identify biomarkers associated with drug sensitivity from pharmacogenomics data of patient-derived cancer organoids. We showed that genes that are proximal to each other in a PPI network are tend to associate with similar patent phenotypes. We devised a method that utilizes both multi-omics data and biological networks to detect robust biomarker genes that successfully identify patients’ responses. We integrated multi-omics data and network-based approaches to identify signature genes to predict recurrence and drug response of cancer patients. We conducted a retrospective study against large scale cohorts that harbor clinical features, transcriptome and tumor mutations. We found that network-based approach outperformed prediction of cancer recurrence and drug response predictions compared to current omics-based approaches. We also discovered that patients with cancer recurrence or without recurrence can be further sub-divided into homogeneous subgroups using signature genes taken from network. Our results indicate the importance of using network biology approaches in cancer genomics.