Cancer Control Research5R03CA097855-02
Daw, Edward W.
A SCORING METHOD FOR MCMC LINKAGE ANALYSIS
DESCRIPTION (provided by applicant): Bayesian Monte Carlo Markov chain (MCMC) techniques have shown promise in dissecting complex genetic traits such as cancer. Cancer is complex both in that failures of more than a single gene are thought to be required to lead to disease and in that both inherited genetic factors and environmental factors play a role. The methods introduced by Heath (1997) and implemented in the program Loki have been able to localize genes contributing to complex traits in both real and simulated data sets. Loki carries out a simultaneous segregation and linkage analysis, estimating not only the location of quantitative trait loci (QTL), but also many other parameters, including the number of QTL, the effects at each QTL, covariate effects (such as environmental exposure), and the segregation patterns of those QTL. These methods can produce posterior probability, distributions for all estimated parameters, and in the past we have focused on the posterior probability distribution of QTL linkage over the genome or simply a particular chromosome to identify regions in which QTL are located. Interpretation of the results of these methods and assessment of their significance has been difficult, meaning that many have found these methods difficult to use and full use of all the information estimated has not been made. We propose to examine a scoring method to produce an easy to interpret score for initial QTL linkage. This score, the Log Of the Posterior placement probability ratio (LOP), designed specifically for complex oligogenic trait linkage detection. LOP contrasts with a lod score in that while a lod score is calculated under a single linkage model, LOP is calculated with Monte Carlo integration over a large number of models. We have done some very promising proof-of-concept work on LOP, but further study is required to completely explore the properties of LOP. We plan to explore how our current implementations of this score perform with different family structures and different trait models. This exploration will result in guidelines for study design and rules for the interpretation of the results, as well as ideas for further improvement of the statistical methods. These guidelines will be used in future studies, which we believe will lead to the identification of additional disease-related genes.