Towards identification of gene interaction networks of human cholesterol biosynthesis

It has long been demonstrated that the level of cholesterol in cells regulatesthe cholesterol biosynthesis through SREBF transcription factors, butlately it has been shown that other factors are also important. To study the system we employed Bayesian network inference and combined it with mathematical modeling and simulation. We constructed a mathematical model of cholesterol biosynthesis and studied its properties through simulation. We measured transcriptional changes of cholesterogenic genes using the Steroltalkmicroarray and treated human hepatocyte samples. We employed Bayesian inference to identify gene-to-gene interactions from both microarray measurements and simulated data. The inferred networks show that the expression of cholesterogenic genes can be predicted from the expression of 4 key genes, one of them being SREBF2. Networks also indicate a strong interaction between SREBF2 and CYP51A1, but not between SREBF2 and HMGCR, the rate-limiting enzyme of cholesterol biosynthesis. The expression of HMGCR seems to be regulated by other factor(s). Computer simulations of the mathematical model of cholesterol biosynthesis exposed that a large number of perturbations of the system is critical for identification of gene-to-gene interactions, and that differences between human individuals (biological variability) and measurement noise (technical variability) pose a serious problem for their automatic inference from DNA microarray data.