[1P-44*] Hidden Markov Modeling of Biomolecular Conformational Dynamics from Atomic Force Microscopy Time-Series Images
High-speed atomic force microscopy (HS-AFM) is a powerful technique for measuring the time-resolved behavior of biomolecules. However, the structural information contained in HS-AFM images is limited to the surface shape of the molecule. Inferring latent three-dimensional structures from HS-AFM images is thus important for getting more insights into the conformational dynamics of the target biomolecule. One of the essential differences between HS-AFM images and other experimental data, such as cryo-electron microscopy, is that HS-AFM images are time-series data, i.e., temporally adjacent images are correlated with each other. Here, to exploit temporal correlations in HS-AFM time-series images and make more accurate estimations of molecular structures, we develop a time-series analysis method based on hidden Markov models. Using simulated HS-AFM image data as a test case, we show that our time-series analysis method makes more accurate estimations of the placement of molecular structure than the estimations from individual images. Furthermore, our method can estimate transition probabilities between representative molecular structures for inferring latent conformational dynamics from HS-AFM images.