Taken together, GeneChip® and BeadArray™ provide the two most widely used SNP chip platforms at the time of writing. We have developed a highly tailored Objective Bayes Hidden-Markov Model (OB-HMM) to automatically infer regions of segmental aneuploidy (copy number variation) from BeadArray™ genotyping data (QuantiSNP). We demonstrate that the Objective Bayes paradigm provides a powerful framework for model building as it affords the benefits of Bayesian marginal probability calculus (information processing) while allowing calibrated hyperparameters in the priors which ensure certain long-running (frequentist) coverage properties (for a general discussion and references on Objective Bayes, see (21,22)). In the context of our work we report on the development of a re-sampling data-driven strategy to automatically set certain prior parameters given a user defined, frequentist, false positive rate. All other parameters are set via maximum marginal likelihood matched to prior training data with known structure. In this way the OB-HMM framework allows for a formal power analysis to be undertaken. Characterization of the power of the method is vitally important in experimental design when sample sizes and end costs are being evaluated.
