Integration of Genetic Variation as External Perturbation to Reverse Engineer Regulatory Networks from Gene Expression Data

Abstract

In systems genetics, genetic variations can be thought as a randomized, multifactorial set of perturbations and the gene/protein expression profile of each individual as the system response to a specific set of perturbations. Current systems genetics approaches, known as genetics genomics, try to combine different types of data such as expression and genetic data both to improve the performance of reverse engineering application and to get deepest biological insights. In this chapter, we present an integrative reverse-engineering approach which exploits both genetic and expression data. The method: 1) codifies genetic-induced perturbations by a variation matrix, which represents differences of mean expressions in each gene according to the genotype of its regulators; 2) define genetic correlation blocks within the variation matrix based on the correlation between genotypes; 3) infers the correct cause-effect pairs based on local peaks of intensity in the variation matrix, since genetic correlation decreases with genetic distance from the real causal gene. Compared to other pair-wise methods typically used in reverse-engineering, the variation matrix shows good performance in terms of both area under receiver operating characteristic and area under the precision versus recall curve. However, on the StatSeq benchmark our approach is able to address a limited number of independent perturbations, due to the high genetic linkage observed in the data. The obtained results provide a basis for advanced integrative approaches able to automate the systematic interpretation of perturbation experiments exploiting the genetic-based prior knowledge.