An integrative genomics approach to the reconstruction of gene networks in segregating populations

Abstract

The reconstruction of genetic networks in mammalian systems is one of the primary goals in biological research, especially as such reconstructions relate to elucidating not only common, polygenic human diseases, but living systems more generally. Here we propose a novel gene network reconstruction algorithm, derived from classic Bayesian network methods, that utilizes naturally occurring genetic variations as a source of perturbations to elucidate the network. This algorithm incorporates relative transcript abundance and genotypic data from segregating populations by employing a generalized scoring function of maximum likelihood commonly used in Bayesian network reconstruction problems. The utility of this novel algorithm is demonstrated via application to liver gene expression data from a segregating mouse population. We demonstrate that the network derived from these data using our novel network reconstruction algorithm is able to capture causal associations between genes that result in increased predictive power, compared to more classically reconstructed networks derived from the same data.

Systematic genetic and genomic analysis of cytochrome P450 enzyme activities in human liver

“…We also acknowledge that one of the limitations of BN approach is that feedback regulation cannot be accommodated. In addition, although we have demonstrated the utility of BNs in predicting how genetic and environmental perturbation signals propagate in biological systems in controlled study populations (Zhu et al 2004(Zhu et al , 2008bSchadt et al 2005;Chen et al 2008;Yang et al 2009), the accuracy to infer causal relationships between genes in a random or uncontrolled study design like the current liver cohort is lower (Wang et al 2009). Furthermore, the directionality of the outcome of any perturbation of genes in a network such as whether the perturbation will raise or lower a phenotypic outcome depends on the genetic background and environmental conditions due to complex network feedbacks (Davey Smith and Ebrahim 2003;Chen et al 2008).…”

“…Unlike coexpression networks, which allow one to look at the overall gene-gene correlation structure at a high level, BNs are sparser but allow a more granular look at the relationships and directional predictions between genes (Zhu et al 2004). Figure 5A (a high-resolution figure is available in Supplemental material) represents an overview of the P450 gene regulatory subnetwork composed of the P450 genes and the known P450 regulators, and the genes that are up to two edges away from them.…”

An integrative genomics approach to infer causal associations between gene expression and disease

“…Microarrays have also helped to identify biomarkers 7 , disease subtypes 3,8,9 and mechanisms of toxicity 10 and, more recently, to elucidate the genetics of gene expression in human populations 11,12 and to reconstruct gene networks by integrating gene-expression and genetic data 13 . The use of molecular profiling technologies as tools to identify genes underlying common, polygenic diseases has been less successful.…”

“…The use of molecular profiling technologies as tools to identify genes underlying common, polygenic diseases has been less successful. Hundreds or even thousands of genes whose expression changes are associated with disease traits have been identified, but determining which of the genes cause disease rather than respond to the disease state has proven difficult.Microarray data have recently been combined with other experimental approaches to facilitate identification of key mechanistic drivers of complex traits 3,[13][14][15][16][17] . One such technique involves treating relative transcript abundances as quantitative traits in segregating populations.…”

Mapping eQTL Networks with Mixed Graphical Markov Models

“…They can be broadly categorized into those extending univariate single-marker regression models to adjust for confounding effects (Leek and Storey 2007;Stegle et al 2010;Listgarten et al 2010) and those using multivariate approaches. The latter can be further categorized into Bayesian networks using conditional mutual information with constraint-based algorithms (Zhu et al 2004), empirical Bayes hierarchical mixtures (Kendziorski et al 2006), directed versions of the PC algorithm (Chaibub Neto et al 2008), structural equation models (Liu et al 2008), sparse partial least squares (Chun and Keleş 2009), fused Lasso regression methods (Kim and Xing 2009), random forests (Michaelson et al 2010), mixed Bayesian networks using the Bayesian information criterion (BIC) with homogeneous conditional Gaussian regression models (Chaibub Neto et al 2010), mixed graphical Markov models restricted to tree network topologies (Edwards et al 2010), sparse factor analysis (Parts et al 2011), and conditional independence tests of order one (Bing and Hoeschele 2005;Chen et al 2007;Kang et al 2010;Chaibub Neto et al 2013).…”