A composite likelihood approach to gaussian network differentiation with application to epigenetics.

For networks originated from dependent populations, methods to test network differentiation between the two populations are generally designed incorporating the nature of dependence. Doing so potentially complicates the inferencing process with heavy computing burden. Through simulations, we assess the value of using composite likelihood to carry out network comparisons under different statuses of population dependency. We apply the method to real-life epigenetic data and assess epigenetic network stability over time.