AbstractConsider the functional interaction of gene A to an interaction subject X; for instance, it is the gene-gene interaction if X represents for a gene, or gene-tissue interaction (expression status) if X for a tissue. In the simplest case, the status of this A-X interaction is r=1 if they are interacted, or r=0 otherwise. A fundamental problem in molecular evolution is, given two homologous (orthologous or paralogous) genes A and B, to what extent their functional overlapping could be by

more »

... e means of interaction networks. Given a set of interaction subjects (X1, ... XN), it is straightforward to calculate the interaction distance (IAB) between genes A and B, by a Markov-chain model. However, since the high throughput interaction data always involve a high level of noises, reliable inference of r=1 or r=0 for each gene remains a big challenge. Consequently, the estimated interaction distance (IAB) is highly sensitive to the cutoff of interaction inference which is subject to some arbitrary. In this paper we will address this issue by developing a statistical method for estimating IAB based on the p-values (significant levels). Computer simulations are carried out to evaluate the performance of different p-value transformations against the uncertainty of interaction networks.