Advances in statistical methods for linkage analysis

The past few years have seen a quantum leap in the amount of genetic marker information available in a typical disease-gene mapping study. Whereas before we were dealing with tens of markers, now we are dealing with hundreds of markers spanning the genome. New computational tools are needed to handle multiple marker analyses efficiently. We describe here a recent advance in statistical methods for linkage analysis that involves the merger of statistics, numerical analysis, computer science, and mathematical genetics: The VITESSE algorithm for rapid exact computation of multipoint lod scores in the presence of untyped individuals. The development of this algorithm involved understanding what parts of the computation were redundant and which parts were essential, and then the creation of a very efficient way to dynamically identify and eliminate the redundant parts. The VITESSE algorithm is much more memory efficient and much faster than previous linkage analysis programs, and significantly extends the size of the problems that can be solved. VITESSE has already proven useful in practice, enabling us to compute better multipoint lod score curves that provide more precise localization of the disease gene. In addition to explaining this algorithm, we discuss briefly our plans for the future, which include parallelization of these algorithms to enable optimal use of modern computational resources.