Models and experimental design for radiation hybrid mapping.

Abstract

Radiation hybrids are a powerful tool for mapping markers on human chromosomes. However, we know little about how we can maximize the mapping power of a hybrid panel and the effects of error on inferences. Further, important situations exist for which current models and model assumptions are inadequate. In this dissertation I examine issues of experimental design and the effects of typing error for radiation hybrids, and develop and implement models for multiple hybrid panels and panels created using selectable markers. First, I compare power of the minimum obligate breaks, maximum likelihood, and Bayesian maximum posterior probability methods for mapping markers using radiation hybrids. I use analytic methods for the tractable case of three loci, and simulation for the multilocus case. I develop guidelines for optimal marker spacing and hybrid sample size for mapping markers and for building framework maps. Second, I consider issues of experimental design, error detection and error correction. Using analytic methods and simulation, I determine the fragment retention rate and ploidy combinations that provide optimal mapping power. I examine the effect of typing error on two-locus information and on ability to correctly map multiple loci. I compare the accuracy of duplicate typed hybrids to that of twice as many single typed hybrids. I develop a recursive algorithm to calculate the distribution of the number of obligate chromosome breaks for a polyploid hybrid, and describe how the algorithm can be used to identify hybrids with suspiciously large numbers of chromosome breaks. Third, I develop two new types of models for radiation hybrids. The first are appropriate when loci typed on two or more independent panels are to be used to build maps. Maps can be built assuming completely independent distances or proportional distances between the panels. The second set of models is for panels created using a selectable marker. The models allow for strong retention of the selectable locus, with retention rates decreasing as we travel farther from the selectable locus in both directions. Both types of models are illustrated with example data.