Top-down phylogenetic tree reconstruction

We propose a novel method for reconstructing phylogenetic trees. It is based on a conceptual clustering method that is an extension of the well-known decision tree learning approach. As such, the method differs from existing methods, both algorithmically and with respect to the information it uses and the assumptions it makes. It scales better in terms of the number of sequences, and as such can be used on large sets of sequences, beyond the point where other methods break down. By nature, it has a stronger tendency to construct trees in which subspecies are defined by particular polymorphisms. Experiments show that its performance is comparable to that of state-of-the-art methods.