Exploiting the association between genomic and gene expression data of natural variants
(patients, different plant species) has already shown to be informative for lead identification, molecular subtyping of patients, biomarker design etc. As naturally occurring genetic variants are not restricted to single or double knock-outs, they contain in principle more information on subtle interactions between molecular entities than classical knock-out experiments. Genotypeexpression phenotype data thus intrinsically are perfectly suited to infer molecular interaction networks. Therefore, in this project a scalable methodology will be developed that not only associates genomic variations to corresponding expression changes (classical eQTL analysis), but that also infers the cellular interaction network underlying the association. Our method is original in (1) maximally exploiting the complementary information contained within publicly available interaction networks and (2) analyzing all complementary data sources (genomic variation, expression variation and network information) in an integrated, rather than a sequential way, which allow us to better cope with the problem of underdetermination from which most eQTL analyses suffer. We will benchmark our methods on publicly available yeast and cancer datasets. To show the ability of our method in inferring causal networks from perturbations triggered by genetic variation in natural variants, we will
analyse a dedicated plant dataset.