Abstract:
|
The identification of gene pathways involved in cancer development and progression and characterization of their activity in terms of multi-platform genomics can provide information leading to discovery of new targeted medications. Such drugs have the potential to be used for precision therapy strategies that personalize treatment based on the biology underlying an individual patient's cancer. We propose a two-step model that integrates multiple genomic platforms, and gene pathway membership information, to efficiently and simultaneously (1) identify genes significantly related to a clinical outcome, (2) identify the genomic platform(s) regulating each important gene, and (3) rank pathways by importance to clinical outcome. We propose a Bayesian model with a novel hierarchical sparsity prior to achieve efficient estimation. Our integrative framework allows us not only to identify important pathways and important genes within pathways, but also to gain insight as to the platform(s) driving the effects mechanistically. We apply our method to a subset of The Cancer Genome Atlas' publicly available glioblastoma multiforme data and identify potential targets for future cancer therapies.
|