Clustered SplitsNetworks

We address the problem of constructing phylogenetic networks using two criteria: the number of cycles and the fit value of the network. Traditionally the fit value is the main objective for evaluating phylogenetic networks. However, a small number of cycles in a network is desired and pointed out in several publications. We propose a new phylogenetic network called CS-network and a method for constructing it. The method is based on the well-known splitstree method. A CS-network contains a face which is k-cycle, k ≥ 3 (not as splitstree). We discuss difficulties of using non-parallelogram faces in splitstree networks. Our method involves clustering and optimization of weights of the network edges. The algorithm for constructing the underlying graph (except the optimization step) has a polynomial time. Experimental results show a good performance of our algorithm.