Faster than neutral evolution of constrained sequences

One of the key insights from the neutral theory of molecular evolution is that functional sequences should generally evolve slower than nonfunctional sequences. The reasoning is very simple - some mutations in functional sequences will damage the function, will be removed by natural selection and thus will not contribute to evolutionary change. It is true that adaptation should speed up evolution because adaptive mutations would contribute to evolution at a much higher probability than neutral ones. This insight is the basis of much of comparative genomics. Find regions that evolve slower than neutral and you find functional sequences even if you know nothing about their function. Find paterns of evolution of functional regions that are too rapid and you find adaptation. Easy. In a paper published in GBE by David Lawrie, Dmitri Petrov and Philipp Messer we show that this is not always so easy. Specifically, when mutation is strongly biased (say in favor of A and T nucleotides) and these nucleotides tend to be weakly deleterious, one can generate very fast flip-flopping between the mutationally preferred state (A and T) and selectively preferred state (G or C in this example). The rate of evolution might even exceed that expected under neutrality without any adaptation. We show that this effect might be important in comparative genomics and urge the development of comparative genomic methods that explicitly incorporate mutational biases, selective processes, and crucially their interactions. The paper has been evaluated by Faculty of 1000 (http://f1000.com/13188956 and here is pdf).