The HMM Classifier. (a) A left-right HMM with Q states. Each state has a transition to itself and outgoing transitions toward higher states behind. Once a state is left it never comes back in a left-right model. (b) Three HMMs are trained separately for promoter, enhancer and background. Log-odds are calculated to classify a genomic region (see Methods).
Won et al. BMC Bioinformatics 2008 9:547 doi:10.1186/1471-2105-9-547