Project

High resolution networks for the analysis of genetic rewiring in cancer

Code
3G045620
Duration
01 January 2020 → 31 December 2023
Funding
Research Foundation - Flanders (FWO)
Research disciplines
  • Natural sciences
    • Bio-informatics
Keywords
genotype phenotype mapping clonal systems alternative splicing network rewiring systems genetics
 
Project description

Network-based methods (NBMs) are promising to identify driver pathways based on cancer systems genetics data NBMs use a network model to search for connected gene sets that are recurrently altered in a cohort of tumor samples These gene sets are proxies for driver pathways and contain next to recurrent drivers also drivers that recur less frequently in a cohort However, current NBMs are sensitive to the overconnectedness of the used network model and therefore require highly curated network priors to drive their analysis Less curated interactions that have the potential to complete missing links or incorporate cancer-specific interactions cannot be accounted for In addition, the too course grained representation of current network models does not allow modeling separately the effects of different mutations in the same gene nor tissue-specific effects, due to alternative splicing We therefore propose a novel NBM based on network representation learning that can convert a more realistic, but also larger and more connected network model into a probabilistic network In this network the edge probabilities reflect how likely the connected nodes interact, taking all different data sources that support an interaction and the local network topology into account Using these network models will allow studying how per patient the effect of putative driver signals propagate in the network and how genetic aberrations can result in cancer-specific network rewiring