Statistical inference is of increasing importance in population genetics as large data sets permit researchers to ask more challenging questions than was feasible in the past. This has profound implications for many important fields such as the understanding of genetic diseases in humans and other species, human evolutionary history, and genetic monitoring of endangered or invasive species. Leading laboratories in this field are all in statistics departments of US and UK universities, whereas most end users of these methods are biologists by training.
The proposer is an experienced biologist with a strong background in statistics. He aims to popularise and improve currently available methods to widen their current audience. His project involves theoretical developments in Monte Carlo methodology as well as computer programming. Both aspects require some training in addition to close interaction with computational statisticians. The training objectives concern mainly the theory of Monte Carlo methods, Importance Sampling and Resembling and pseudo-likelihood based methods.
The project is aimed at widening the scope of methods of inference to a broad set of population evolutionary scenarios including divergence, admixture and variable size. Two complementary approaches will be pursued: one based of the estimation of likelihood and the other on Approximate Bayesian Computation (ABC). Likelihood based approaches will involve Importance Sampling schemes, Markov Chain Monte Carlo (including population methods) and Sampling Importance Resembling. Improvement of the efficiency of the ABC approach is expected from exploiting adaptive sampling of parameters. For both approaches, generic software will be developed to widen their appeal to a broader audience. The realization of this project will finally benefit to many population geneticists who will dispose of a versatile and simple to use software to analyse more thoroughly their molecular datasets.
Fields of science
Call for proposal
See other projects for this call