Leveraging comprehensive cancer systems genetic data to uncover the modus operandi of driver mutations
During life our DNA undergoes many somatic aberrations of which only a small fraction effectively
enable cells to become cancerous (drivers). Methods that rely on a prior interaction network have
delivered promising results in identifying these drivers. However, current models highly
oversimplify biological reality: The networks that drive the analysis do not capture the context of
the tumor. Also, because aberrations are modeled at gene level the possibility that distinct
mutations in the same gene affect this gene differently across samples cannot be considered. To
cope with these problems, we will 1) use a data-driven approach to construct a context-specific
probabilistic network and 2) develop a model that allows exploiting the yet unexploited functional
information to trace back the effect of each single aberration over this network. To do so, the
method assumes that a driver aberration elicits a response that propagates through the network.
Key to the method is the use of an aberrant expression phenotype (differential expression level,
isoform switches, U+2026) to trace and quantify this propagating response, called the path of influence
(POI). Based on characteristics of their POI, each somatic aberration is scored to identify drivers.
Cohort-level analysis then allows identifying driver pathways by searching for overrepresented
POIs in high-scoring aberrations. Applying the method to the ICGC PAWG data will contribute to a
systems level understanding of tumorigenesis.