Hidden Markov Models

Kenji Okamoto

David R. Westhead, M. S. Vijayabaskar

Structural dynamics of biomolecules, such as proteins, plays essential roles in many biological phenomena at molecular level. It is crucial to understand such dynamics in recent biology. The single-molecule Förster resonance energy transfer (smFRET) measurement is one of few methods that enable us to observe structural changes of biomolecules in realtime. Time series data of smFRET, however, typically contains significant fluctuation, making analysis difficult. On the other hand, one can often assume a Markov process behind such data so that the hidden Markov model (HMM) can be used to reproduce a state transition trajectory (STT). The common solution of the HMM can be used for smFRET data, too, while one has to define the specific model, i.e., the observable variable and the emission probability. There are several choices of the model for smFRET depending on the measurement method, the detector type, and so on. I introduce some of applicable models for smFRET time series data analysis.