Summarizing multiple gene trees using cluster networks

Abstract

The result of a multiple gene tree analysis is usually a number of different tree topologies that are each supported by a significant proportion of the genes. We introduce the concept of a cluster network that can be used to combine such trees into a single rooted network, which can be drawn either as a cladogram or phylogram. In contrast to split networks, which can grow exponentially in the size of the input, cluster networks grow only quadratically. A cluster network is easily computed using a modification of the tree-popping algorithm, which we call network-popping. The approach has been implemented as part of the Dendroscope tree-drawing program and its application is illustrated using data and results from three recent studies on large numbers of gene trees. © 2008 Springer-Verlag Berlin Heidelberg.