MOTIVATION: Empirical analyses of protein secondary structures based on circular dichroism (CD) and synchrotron radiation circular dichroism (SRCD) spectroscopic data rely on the availability of reference datasets comprised of spectra of relevant proteins, whose crystal structures have been determined. Datasets comprised of only soluble proteins have not proven suitable for analysing the spectra of membrane proteins. RESULTS: A new reference dataset, MP180, has been created containing the spectra of 30 membrane proteins encompassing the secondary structure and fold space covered by all known membrane protein structures. In addition a mixed soluble and membrane protein dataset, SMP180, has been created, which includes 98 soluble protein spectra (SP) plus the MP180 spectra. Calculations of both membrane and soluble protein secondary structures using SMP180 are significantly improved with respect to those produced, using soluble protein-only datasets. The SMP180 dataset also enables determination of the percentage of transmembrane residues, thus enhancing the information previously obtainable from CD spectroscopy. AVAILABILITY AND IMPLEMENTATION: Reference dataset online at the DichroWeb analysis server (http://dichroweb.cryst.bbk.ac.uk); individual protein spectra in the Protein Circular Dichroism Data Bank (http://pcddb.cryst.bbk.ac.uk).

MOTIVATION: Circular Dichroism (CD) spectroscopy is a long-established technique for studying protein secondary structures in solution. Empirical analyses of CD data rely on the availability of reference datasets comprised of far-UV CD spectra of proteins whose crystal structures have been determined. This article reports on the creation of a new reference dataset which effectively covers both secondary structure and fold space, and uses the higher information content available in synchrotron radiation circular dichroism (SRCD) spectra to more accurately predict secondary structure than has been possible with existing reference datasets. It also examines the effects of wavelength range, structural redundancy and different means of categorizing secondary structures on the accuracy of the analyses. In addition, it describes a novel use of hierarchical cluster analyses to identify protein relatedness based on spectral properties alone. The databases are shown to be applicable in both conventional CD and SRCD spectroscopic analyses of proteins. Hence, by combining new bioinformatics and biophysical methods, a database has been produced that should have wide applicability as a tool for structural molecular biology.

BACKGROUND: Circular Dichroism (CD) spectroscopy is a widely used method for studying protein structures in solution. Modern synchrotron radiation CD (SRCD) instruments have considerably higher photon fluxes than do conventional lab-based CD instruments, and hence have the ability to routinely measure CD data to much lower wavelengths. Recently a new reference dataset of SRCD spectra of proteins of known structure, designed to cover secondary structure and fold space, has been produced which includes low wavelength (vacuum ultraviolet - VUV) data. However, the existing algorithms used to calculate protein secondary structures from CD data have not been designed to take optimal advantage of the additional information in these low wavelength data. RESULTS: In this study, we have optimised secondary structure calculation methods based on the low wavelength CD data by examining existing algorithms and secondary structure assignment schemes, and then developing new methods which have produced clear improvements in prediction accuracy, especially for beta-sheet components. We have further shown that if precise measurements of protein concentrations, and therefore spectral magnitudes, are not available, the inclusion of the low wavelength data will significantly improve the analyses. However, we have also demonstrated that the new reference dataset, methods, and assignments can also improve the analyses of conventional circular dichroism data, even if the low wavelength data is not available. CONCLUSION: VUV CD data include important information on protein structure which can be exploited with the algorithms and methodologies described.