Bayesian inference of accurate population sizes and FRET efficiencies from single diffusing biomolecules.

Rebecca R Murphy,George Danezis,Mathew H Horrocks,Sophie E Jackson,David Klenerman

doi:10.1021/ac501188r

Rebecca R Murphy, George Danezis + Show 3 more

Open Access

https://doi.org/10.1021/ac501188r

Copy DOI

Abstract

It is of significant biophysical interest to obtain accurate intramolecular distance information and population sizes from single-molecule Förster resonance energy transfer (smFRET) data obtained from biomolecules in solution. Experimental methods of increasing cost and complexity are being developed to improve the accuracy and precision of data collection. However, the analysis of smFRET data sets currently relies on simplistic, and often arbitrary methods, for the selection and denoising of fluorescent bursts. Although these methods are satisfactory for the analysis of simple, low-noise systems with intermediate FRET efficiencies, they display systematic inaccuracies when applied to more complex systems. We have developed an inference method for the analysis of smFRET data from solution studies based on rigorous model-based Bayesian techniques. We implement a Monte Carlo Markov chain (MCMC) based algorithm that simultaneously estimates population sizes and intramolecular distance information directly from a raw smFRET data set, with no intermediate event selection and denoising steps. Here, we present both our parametric model of the smFRET process and the algorithm developed for data analysis. We test the algorithm using a combination of simulated data sets and data from dual-labeled DNA molecules. We demonstrate that our model-based method systematically outperforms threshold-based techniques in accurately inferring both population sizes and intramolecular distances.

Highlights

Förster resonance energy transfer (FRET) is a nonradiative energy transfer from a donor (D) to an acceptor fluorophore (A), where the efficiency of energy transfer depends on their separation, r ( ) E =
In a diffusion based singlemolecule Förster resonance energy transfer (smFRET) experiment, photons emitted from the donor and acceptor fluorophores are collected in a continuous stream and time binned on a time scale comparable with the average dwell time of a molecule diffusing through the confocal volume
Assuming a bin time of 1 ms, a typical experimental data set (10−20 min of data) would include 6−12 million bins. It is a significant achievement of the inference method that it makes extremely accurate estimates of the FRET efficiency using only 100 000 bins, corresponding to less than 2 min of data

Summary

Introduction

Determining intramolecular distance information and population sizes from smFRET experiments, remains challenging,[11] as incomplete sample labeling, photophysical artifacts, unequal photon detection, and the stochastic nature of diffusion through the confocal volume[11] as well as linker dynamics[12] hamper development of quantitative smFRET experiments. Using smFRET data to constrain molecular dynamics simulations can improve accuracy of structural information.[13,14] Linear flow has been used to reduce heterogeneity in confocal dwell time and diffusion pathway.[15,16] Methods to determine correction factors,[17] development of alternating-laser excitation (ALEX) techniques[18−21] and multiparameter fluorescence detection. Article (MFD)[22] as well as more sophisticated burst-selection algorithms[11,23] allow more accurate identification of fluorescent bursts

Methods

Results

Conclusion