Recently, Microarry technology has been widely applied in the biomedical researches, and the potential usefulness of gene expression profiles for tumor sample classifications, prediction of genetic regulatory systems and the drug target effect by different treatments have been shown. This poster introduces graphical models as a natural environment in which to formulate and solve problems in Microarray data analysis. Especially emphasis is given to the prediction of gene regulatory network by Graphical Gaussian model (GGM) and quantify the effects of different experiment treatment on gene expression profiles by Graphical Log-line model (GLM). The power of graphical models is explored and illustrated through two example applications.
By using our newly developed causal prediction program MGraph, four MAPK pathways in yeast, pheromone response pathway, protein kinase C (PKC) pathway, high osmolarity (HOG) pathway and filamentous growth pathway have been reasonably reconstructed through GGM. The prediction of MAP Kinase signaling networks with the missing value replaced by zero and missing value estimated by K-nearest neighbors are compared. GLM is used to quantify the contributions of the sex, genotype and age to transcriptional variance in Drosophila melanogaster, our predicted causal effects are consistent with the original biological results.