Correcting the apparent mutation rate acceleration at shorter time scales under a Jukes-Cantor model
At macroevolutionary time scales, and for a constant mutation rate, there is an expected linear relationship between time and the number of inferred neutral mutations (the "molecular clock"). However, at shorter time scales a number of recent studies have observed an apparent acceleration in the rate of molecular evolution. We study this apparent acceleration under a Jukes-Cantor model applied to a randomly mating population, and show that, under the model, it arises as a consequence of ignoring short term effects due to existing diversity within the population. The acceleration can be accounted for by adding the correction term h_0e^{-4mu*t/3} to the usual Jukes-Cantor formula p(t)=(3/4)(1-e^{-4mu*t/3}), where h_0 is the expected heterozygosity in the population at time t=0. The true mutation rate mu may then be recovered, even if h_0 is not known, by estimating mu and h_0 simultaneously using least squares. Rate estimates made without the correction term (that is, incorrectly assuming the population to be homogeneous) will result in a divergent rate curve of the form mu_{div}=mu+C/t, so that the mutation rate appears to approach infinity as the time scale approaches zero. While our quantitative results apply only to the Jukes-Cantor model, it is reasonable to suppose that the qualitative picture that emerges also applies to more complex models. Our study therefore demonstrates the importance of properly accounting for any ancestral diversity, as it may otherwise play a dominant role in rate overestimation.