Estimating mutational pathways of carcinogenesis
by
Moritz Gerstung
Dep. of Biosystems Science and Engineering, ETH Zurich
Coauthors: Seth Sullivant and Niko Beerenwinkel

The development of cancer is thought to be a multistep process of accumulating mutations. We describe this process by an exponential waiting time model in which the chronology of events is constrained by a partial order. This mathematical structure defines the conjunctive Bayesian network, a probabilistic model accounting for multiple pathways of carcinogenesis. We introduce a censored version of this model and estimate model parameters and the underlying partial order from observed mutation data by an EM algorithm. The expected waiting time of a given mutational pattern measures the genetic progression of the tumor and presents a marker of disease progression. We apply our model to comparative genomic hybridization (CGH) data from different cancer types revealing characteristic mutagenetic pathways. Furthermore we evaluate the prognostic value of the genetic progression score using patient survival data.

Date received: May 16, 2008