A Comparative Study of Multivariate and Univariate Hidden Markov Modelings in Time-Binned Single-Molecule FRET Data Analysis

We compare two different types of hidden Markov modeling (HMM) algorithms, e.g., multivariate HMM (MHMM) and univariate HMM (UHMM), for the analysis of time-binned single-molecule fluorescence energy transfer (smFRET) data. In MHMM, the original two channel signals, i.e., the donor fluorescence intensity (I(D)) and acceptor fluorescence intensity (I(A)), are simultaneously analyzed. However, in UHMM, only the calculated FRET trajectory is analyzed. On the basis of the analysis of both synthetic and experimental data, we find that, if the noise in the signal is described with a proper probability distribution, MHMM generally outperforms UHMM. We also show that, in the case of multiple trajectories, analyzing them simultaneously gives better results than averaging over individual analysis results.