Implementation of a Bayesian Secondary Structure Estimation Method for the SESCA Circular Dichroism Analysis Package

Circular dichroism spectroscopy is a structural biology technique frequently applied to determine the secondary structure composition of soluble proteins. Our recently introduced computational analysis package SESCA aids the interpretation of protein circular dichroism spectra and enables the validation of proposed corresponding structural models. To further these aims, we present the implementation and characterization of a new Bayesian secondary structure estimation method in SESCA, termed SESCA_bayes. SESCA_bayes samples possible secondary structures using a Monte Carlo scheme, driven by the likelihood of estimated scaling errors and non-secondary-structure contributions of the measured spectrum. SESCA_bayes provides an estimated secondary structure composition and separate uncertainties on the fraction of residues in each secondary structure class. It also assists efficient model validation by providing a posterior secondary structure probability distribution based on the measured spectrum. Our presented study indicates that SESCA_bayes estimates the secondary structure composition with a significantly smaller uncertainty than its predecessor, SESCA_deconv, which is based on spectrum deconvolution. Further, the mean accuracy of the two methods in our analysis is comparable, but SESCA_bayes provides more accurate estimates for circular dichroism spectra that contain considerable non-SS contributions.

BAYESIAN SPECTRUM DECONVOLUTION INCLUDING UNCERTAINTIES AND MODEL SELECTION: APPLICATION TO X-RAY EMISSION DATA USING WINBUGS

Spectrum deconvolution is an important task in ionizing radiation measurements, as the pulse height spectra, or, in general, the measured data from spectrometers or other measuring instruments are usually determined by the convolution of the response function with the fluence spectra. The method presented here for obtaining fluence spectra from the measurements is an application of Bayesian parameter estimation to the deconvolution of X-ray emission data. The problem of choosing the optimal model among several possible models is also considered, as well as an approach to include contributions from various sources of uncertainty, both correlated and uncorrelated. The application is carried out using the Bayesian software WinBUGS.