Microarray is a technique which could measure thousands of genes’ expression level simultaneously. To find the co-regulated genes in microarray data, the clustering algorithms have been applied, such as hierarchical, K-means and SOM. Numerous different clustering results have been produced. It is a big challenge for biologists to choose meaningful clusters among the huge amount of results. The quantitative measurement of clustering result is called cluster validation. The cluster validation could be divided into two methods: data-driven approach and knowledge-driven approach based on the distance measurement between genes. We propose a new information fusion based distance metrics which could combine two knowledge information: data information and prior biological knowledge. And firstly incorporating the database of interacting protein to deal with the uncertainty of prior knowledge and using the optimization methods to find the optimal parameters for information fusion equation. To check the effect of the new method, two test datasets are used for experiments. In the comparison with conventional distance measurements, the new method shows better performance.