Towards Optimal Distance Functions for Stochastic Substitutions Models

This talk concentrates on the task of inter-species distance estimation, which prior to this work was, by and large, assumed to be implied by the given model of evolution. We show that for most known models, the set of valid distance functions form a linear space. This implies that it is possible to adjust to each given set of aligned DNA sequences a different distance function. We show that a deliberate selection of the function used may significantly improve the accuracy of distance estimates and, consequently, also the accuracy of the reconstructed tree.
Our theoretical analysis is supported by reconstructions of trees in the Kimmura two parameters model.
The talk is self-contained, and does not assume any prior knowledge on phylogenetic reconstruction.