MALANI to find "Class II cancer genes"
Class I cancer genes are those frequently mutated oncogenes and/or genes that are differentially expressed under cancerous contexts. The Class II cancer genes are genes that are neither differentially expressed nor mutated but act as invisible hands “connecting the dots” to complete oncogenic signals in a cancer network.
Cancer networks are not restricted to mutated and differentially expressed genes. We hypothesize the presence of “Class II cancer genes” which are often disregarded due to no detected mutation or differential expression, but are important in coordinating the functionality in cancer networks. However, hunting Class II cancer genes out of sea of irrelevant genes is a daunting task. Here, we harness the powers on machine learning and network biology using 8807 expression arrays correspond to 9 cancer types, with more than 200*10^6 Support Vector Machine (SVM) models built to reconstruct a cancer network. Our results indicate less than 3% out of ~19,000 not differentially expressed genes are Class II cancer gene candidates. Using genome-wide mutational data, our results suggest the existence of Class II cancer genes that coordinate cancer network functionality such as SLC19A1 and ATAD3B were recently reported to associate to breast cancer outcomes. We therefore have provided a powerful tool to hunt Class II cancer genes from cancer networks. More importantly, due to rare mutational events in Class II cancer genes, our work provides a rationale to develop Class II cancer gene-based therapeutics with a much stable shelf life in clinics.