Glimpses from the strange world of phylogenetic mixtures
Seminar Room 1, Newton Institute
Rates of evolution and evolutionary history may vary across the genome of a given organism. The standard way of modeling the resulting data is a "mixture model" where a given phylogenetic pattern comes from one of a collection of trees with fixed probabilities. In this talk I will focus on recent results (joint with Mike Steel and Elchanan Mossel) which investigate what pattern probabilities are and are not possible when rates of evolution vary on a single topology.
In particular, we have shown that a mixture of two processes on a tree of one topology can produce exactly the same expected site pattern frequencies (i.e. sequence data) as an unmixed process on a different topology. For mixtures of a number of processes the situation is even more dire: in a certain sense the topology is not uniquely determined for a majority of data vectors. I will also present other related results along with the associated mathematics and indicate some interesting future directions for research.
If it doesn't, something may have gone wrong with our embedded player.
We'll get it fixed as soon as possible.