|
Abstract:
|
There is growing evidence that large-scale copy number polymorphisms (CNPs) contribute significantly more to human genetic variation and diversity than recently thought. Variations in CNPs have been implicated in a variety of disease phenotypes, especially cancers. Comparative genomic hybridization (CGH) arrays simultaneously assay tens of thousands of loci and are used to infer genome-wide copy number profiles. Like all genomic microarray tiling technologies, however, CGH arrays are subject to measurement error and other inconsistencies. We propose an empirical Bayes approach to infer copy number profiles from CGH arrays. The model structure enables the sharing of power between loci, particularly in estimating measurement error, and incorporates a spatial structure to facilitate the identification of CNPs which span multiple CGH loci.
|
