Jaromir Sant (University of Oxford)

Inferring natural selection and allele age from allele frequency time series data via exact simulation

Abstract. A standard problem in population genetics is that of inferring evolutionary and biological parameters such as the strength of natural selection and allele age from DNA samples. Up until recently, most of the datasets available for such a task were obtained by sampling individuals from a contemporaneous population simultaneously, thereby offering a static snapshot of the population’s genetic diversity. However, recent advances in DNA extraction and sequencing technologies have allowed for ancient DNA found at archaeological sites to be analysed, leading to the creation of several time-series datasets detailing historical changes in allele frequencies. These datasets have paved the way for the development of powerful inferential techniques which explicitly exploit this new temporal dependence in the data to provide better estimators. In this talk I shall introduce a Markov Chain Monte Carlo framework which allows for inferring selection and allele age based on time-series data coming from an underlying Wright-Fisher diffusion. The chief novelty is that by augmenting the state space with the unobserved diffusion trajectory we are able to develop an efficient method in which trajectory updates and accept/reject probabilities can be calculated without error, in spite of the diffusion’s intractable transition density. We illustrate the method’s performance on simulated data, and subsequently apply it to an ancient DNA horse dataset. This is joint work with Paul Jenkins (Warwick), Jere Koskela (Newcastle), and Dario Spanò (Warwick).