|
Martin Zacharias, Professor of Computational Biology, and his group submitted the best computational interaction model in the latest challenge of Critical Assessment of PRedicted Interaction (CAPRI) between two protein molecules out of a total of 350 submissions from scientist worldwide. With an average deviation of ~ 0.25 nanometres of the predicted atom positions from the actual, experimentally determined atom positions at the interface of the two interacting target molecules the Jacobs scientists achieved a prediction accuracy very close to the resolution of experimentally determined complex structures.
[
Nov
23, 2007]
Almost all biological processes involve protein-protein interactions. A detailed understanding of the function of these protein-protein interactions is very valuable, not least because it ultimately elucidates medical possibilities of influencing these interactions by specifically designed drug molecules. This requires the knowledge of the three-dimensional structure of the protein-protein complex. The experimental determination of all complex structures is not possible due to the time and costs involved. Therefore the realistic prediction of protein-protein complexes (protein-protein docking) is of great importance, and the development and improvement of virtual docking methods to achieve realistic predictions is a rapidly evolving field.
The CAPRI (C ritical A ssessment of PR edicted I nteraction) challenge
The prediction of protein complex structures is achieved by computer programs that evaluate all the information available on atomic interactions that are determined by the amino acid sequence of the proteins. Comparable to a complicated puzzle, where one tries to identify the correct mode of protein binding out of a myriad of alternative arrangements the computer than tries to find the model best suited to account for all conformational changes – general as well as local – during the simulation of the binding process. The CAPRI challenge, which has been hosted by the European Bioinformatics Institute of the European Molecular Biology Laboratory (EMBL) once or twice per year since 2001, offers a community-wide opportunity to evaluate and compare different methods and protocols for protein-protein docking: Based on the known structure of the component proteins about 40 participating research groups submit predicted protein-protein complex structures shortly before the complex structure is experimentally determined by X-ray crystallography. Upon availability of the experimental complex structure, predictions are evaluated by direct comparison with experiment. Usually the participants have 2 - 3 weeks for working out their simulations.
In the most recent CAPRI round 14, of which the results were published last week,
Martin Zacharias and his group used the protein docking computer program ATTRACT. Developed by the group itself and first published in 2003, the ATTRACT approach employs a reduced protein model in combination with computationally very efficient methods to account for local and global conformational changes during simulation of complex formation. “The predicted protein-protein interaction plays an important role in the regulation of cell adhesion, which is of major importance during the growth of multi-cellular organisms and also for understanding the development of cancer,” says computational biologist Zacharias about the target of the CAPRI round 14. “The target was so complex, that the whole challenge with hundreds of submissions only generated two acceptable predictions. We are very proud, that our prediction not only was the best, but almost twice as good in accuracy as the second best prediction. Compared to a resolution of about 0.15 nanometers of molecule structures determined experimentally by X-ray crystallography our atom position deviation average of 0.25 nm is a big step forward in prediction quality,” comments Zacharias on the challenge result.
|
