Department of Computer Engineering
CS 590 SEMINAR
MEMNAR: Finding Mutually Exclusive Mutations through Negative Association Rule Mining
Computer Engineering Department
Cancer is a heterogeneous disease; where different individuals harbor different combinations of genomic alterations. Of these alterations only a small set of mutations drive cancer development while many of them are passengers with no significant consequences on cancer. As large cohorts of cancer patients sequenced in multiple cancer types, it has been observed that genes mutation do not co-occur in the same tumor. As a mutual exclusivity relationship is due to a functional relationship between the genes, the mutual exclusive gene sets can help reveal cancer-driving mechanisms. We address the problem of discovering mutually exclusive mutation gene sets in cancer through mining negative association rules. We developed an algorithm, MEMNAR, which efficiently mines for negative association rules in patient mutation data and construct gene sets based on these extracted rules. MEMNAR can also detect complex patterns that are not possible with the current approaches. Synthetic data experiments show that MEMNAR can discover mutual exclusive gene sets faster and with improved accuracy. When we apply MEMNAR on 12 cancer types from the PanCancer genome project, we identify several mutually exclusive gene sets that are biologically relevant. When the common gene sets between different cancer types are analyzed, certain cancers are found to share a large number of mutual exclusive mutations. MEMNAR holds potential to shed light on the undiscovered mechanisms of cancer and aid developing targeted therapies.
DATE: 24 April, 2017, Monday @ 16:55