The java program for performing the significance test for conserved genomic regions can be found here.
The computation of the p-value of a given gene cluster is based on a compound Poisson approximation.
For more details, see:
Grusea, S., Pardoux, P., Chabrol, O. and Pontarotti, P. (2011), "Compound Poisson approximation and testing for gene clusters with multigene families", Journal of Computational Biology, vol. 18(4), pp. 579-594.