X-ray crystallography is the predominant method for obtaining three dimensional atomic information for proteins. Determining the three dimensional protein crystal structure from weak data can fail with current computational methods. We have developed a novel algorithm to solve structures from data with a weak signal and push the limits of X-ray crystallography.
Experimental methods, such as X-ray crystallography and electron microscopy, require computational methods to obtain an accurate three dimensional representation. The structure solution process is difficult and fails in cases when only data with a weak signal can be obtained. In particular, weak data from large macromolecular complexes and membrane proteins of considerable medical interest may not allow structure solution.
Currently, structure determination relies on a multi-step approach with successive approximations of the experimental data in each step. We developed a method that combine all the experimental data with all the relevant prior information in one step.
Novel algorithms allow structure solution when current methods fail. For example, the application of the our new algorithm to low resolution X-ray diffraction data from RNA polymerase II crystals has led to an automatically built molecule when current methods fail. The Figure shown above is the electron density of a portion of a molecule shown in blue and the automatically built model is multicoloured.
The research will meet the urgent need to solve macromolecular structures from data with a weak signal in software that seemlessly automates the process and is ideal for non-specialists.
Currently, the process of solving a macromolecular crystal structure consists of distinct steps. First, an initial electron density is constructed from the X-ray data in “experimental phasing”. Then, expected features of macromolecular electron density, such as the flatness of solvent regions, are imposed on the experimental electron density to improve its quality. This “density modified” map is typically combined with the initial experimental density map in “phase combination”. Finally, the resulting electron density is used to iteratively build and refine a model of the macromolecule. We have developed a novel "combined” approach (shown in the flow chart) that directly considers phase
information from the experimentally collected X-ray data and simultaneously combines it with the information from density modification and model building into a single unified process. Thus, the structure solution process no longer relies on successive step-wise approximations of the experimental data.
Skubak and Pannu (2013) Automatic protein structure solution from weak X-ray data.
Nature Communications 4, 2777.