Abstract
Complex traits may be determined by the expression levels of a number of genes and their interactions. Microarray data enables high throughput approach to measure the correlations between transcripts and phenotypes. Biologically meaningful message may be obtained by correlations between the phenotypes and the sets of known pathways. Here, we propose a likelihood-based graphical modeling approach to get the system-biological groundview of molecular mechanism behind the variation of phenotype. The geometry of the graph is selected based on AIC. From the local structure around the phenotype-node, it is possible to estimate the direct effect. By tracing a series of the hubs from the phenotype-node towards the center of the graph, it is possible to estimate the hierarchical module structure controlling the phenotype.
We applied our method to a public data of sockeye salmon (Oncorhynchus nerka) and estimate the gene expression network. A total of 80 individuals of sockeye salmon were captured following their spawning migration stages, and expressions of 16006 genes were measured using white muscle tissues.Estimated gene network shows that sexual hormones drive the spawning migration, and the environmental stress in the process of migration activates glycolysis and muscle growth. By integrating the phenotype, expressions and marker genotype data, it is possible to quantify the direct and indirect causal effect of genetic perturbations.Our method will elucidate the mechanisms of the genetic effects of hatchery supplementation and aquaculture on wild populations.
Reiichiro Nakamichi
Tokyo University of Marine Science and Technology
Further reading
Directed graphical modeling of gene expression profile underlying salmonids reproductive behavior (pdf, 8 pages) by Reiichiro Nakamichi, Hirohisa Kishino and Shuichi Kitada.