Bayesian and maximum likelihood estimation of genetic maps
Thomas L. York, Richard Durrett, Steven Tanksley and Rasmus Nielsen
Abstract. There has recently been increased interest in the use of Markov Chain Monte Carlo (MCMC) based Bayesian methods for estimating genetic maps. The advantage of these methods is that they accurately can deal with missing data and genotyping errors. In this paper we present an extension of the previous methods that makes the Bayesian method applicable to large data sets, and an extensive simulation study examining the statistical properties of the method and comparing it to the likelihood method implemented in MAPMAKER. We show that the Maximum A Posteriori (MAP) estimator of the genetic distances, corresponding to the maximum likelihood estimator, performs better than estimators based on the posterior expectation. We also show that while the performance is similar between MAPMAKER and the MCMC based method in the absence of genotyping errors, the MCMC based method has a distinct advantage in the presence of genotyping errors. A similar advantage of the Bayesian method was not observed for missing data. We also reanalyze a recently published set of data from the eggplant and show that the use of the MCMC based method leads to smaller estimates of genetic distances.
Preprint as PDF
Back to Durrett's home page