Stanford University - Departments of Statistics & Health Research and Policy
Radiation hybrids are a popular method of creating a stochastic set of experimental data for mapping DNA markers. In this talk, a new method for constructing maps from radiation hybrid data is proposed. It uses maximization of a pseudo-likelihood based on information from all marker pairs. The pseudo-likelihood is a relatively smooth function of the marker positions, allowing us to use gradient-based computational tools. Multi-dimensional scaling is used to provide starting positions for the markers. The method is fully automatic, making it possible to find a good map for a hundred markers in a few minutes.
Information from anchor markers can also be incorporated in a natural way.
This is joint work with Laura Lazzeroni, Trevor Hastie, Adam Olshen, & David Cox.