Abstract
Fluorescence spectroscopic and imaging techniques, such as fluorescence-correlation spectroscopy, image correlation spectroscopy, time-resolved fluorescence spectroscopy, and intensity-based spectroscopy, can provide sparse time-dependent positional and inter-fluorophore distance information for macromolecules and their complexes in vitro and in living cells. Here, we formulated a Bayesian framework for processing and using the fluorescence data for interpreting by static and dynamic models of biomolecules. We introduce Bayesian Fluorescence Framework (BFF) as part of the open-source Integrative Modeling Platform (IMP) software environment, facilitating the development of modeling protocols based in part on fluorescence data. BFF improves the accuracy, precision, and completeness of the resulting models by formulating the modeling problem as a sampling problem dependent on general and flexible libraries of (i) atomic and coarse-grained molecular representations of single-state models, multi-state models, and dynamic processes, (ii) Bayesian data likelihoods and priors, as well as (iii) sampling schemes. To illustrate the framework, we apply it to a sample synthetic single-molecule FRET dataset of the human transglutaminase 2. We show how to integrate time-resolved fluorescence intensities, fluorescence correlation spectroscopy curves, and fluorescence anisotropies to simultaneously resolve dynamic structures, state populations, and molecular kinetics. As BFF is part of IMP, fluorescence data can be easily integrated with other data types to solve challenging modeling problems.
Statement of Significance Bayesian Framework for Fluorescence (BFF) is software that implements a probabilistic framework for processing experimental fluorescence data to provide input information for Bayesian integrative structure modeling. BFF facilitates constructing integrative modeling protocols based in part on fluorescence data by reducing the required fluorescence spectroscopy and microscopy domain knowledge. In addition, it improves the precision and accuracy of the resulting models.
Competing Interest Statement
The authors have declared no competing interest.