We believe our approach is the first application of OB-HMM to high-throughput genomic datasets. In genomic data analysis using HMMs it is often the case that one of the hidden states carries special status as a ‘null’ or normal state. In this scenario, we believe the OB framework provides a powerful approach which allows for calibrated Type I error rates of excursions out of the null state, while affording the benefits of marginal probability calculus that defines the Bayesian approach.
