Genome maps play an essential role in human genetic research. They provide known frameworks for locating disease genes and form a scaffold for genome-wide sequencing. Radiation hybrid mapping is a technique for creating human genome maps of arbitrary resolution (depending on the radiation dosage used). Thus, radiation hybrid maps can provide the cohesion necessary for integrating several different types of mapping data.
In this talk, we discuss methods for constructing genome-wide radiation hybrid maps, and we introduce RHMAPPER, an interactive computer package incorporating our techniques. The package includes facilities that automatically detect and flag errors in the data, allowing for error correction during map assembly. We describe the Hidden Markov Model used to represent uncertainty in the data, and we present algorithms for finding good marker orders from the exponentially-large space of possible orders. Our methods can handle diploid as well as haploid data and are efficient enough to build maps with hundreds or thousands of markers.
We describe how we used these techniques to build and evaluate a radiation hybrid map of over 11,000 STSs, with estimated 99% coverage of the human genome. We have combined this map with STS-content data and the Genethon genetic map to produce an integrated genome-wide map of over 20,000 markers, two-thirds the number required for the U.S. Human Genome Project's 1998 goal.