Using Reversible Jump-MCMC to detect multiple quantitative trait loci and their effects in crops

Identifying likely locations of genes, termed quantitative trait loci or QTL, that affect traits of interest in experimental designs is important in modern agriculture.

This talk gives a brief history of the statistical methodologies that have been employed, and outlines a flexible Bayesian model framework that can estimate the number, location and size of effect of these putative QTL. Since the Bayesian posterior has a complex form, sampling of the posterior is performed using Reversible jump MCMC (RJ-MCMC). Initialization of the chain is examined, and a new RJ-MCMC step - termed an exchange - is introduced.

Bayes factors are examined as a means of model selection. In the absence of good prior information about the number of QTL, this seems prudent since the posterior distribution for the number of QTL is greatly shaped by the prior. Efficient estimation of the Bayes factor is also discussed.