Abstract
Previous studies of genetic interactions have focused on measuring and analyzing the level of mRNA expressed by each gene and then finding statistical correlations between them. Here we focus on assessing the global pattern, not the level of activity of specific genes. We formulate models of different architectures of genetic interactions, including random and scale free patterns with homogeneous, heterogeneous, and symmetric connection topologies. We then compute the probability density function (PDF) of the mRNA levels produced by each model. Thus, we make a dictionary of the mRNA patterns produced by different architectures of genetic interactions. Then we compare these results to the statistics of mRNA levels experimentally measured by cDNA microarrays. We then read our dictionary backwards, starting from the statistical patterns of the mRNA levels, to find the pattern of genetic interactions responsible for the observed experimental data. We find that the data is best represented by a model where different genes have different patterns of input regulation from the other genes but the same patterns of output regulation to the other genes.
