| Contact information|
dry-lab: Z5-3031 - Tel: +1-819-820-6868 ext. 12472
The study of protein function is a key step in the comprehension of disease mechanisms and their treatment. While protein function is closely related to dynamics, most known protein structures have been elucidated by X ray diffraction, which gives little information about movements (in the form of b-factors). The static representation of proteins given by X-ray crystallography fails to represent such movements. In reality proteins are dynamic entities and instead of a single conformation, proteins should be considered as an ensemble of conformations.
Two main techniques are principally used to evaluate the dynamic nature of proteins: nuclear magnetic resonance and molecular dynamic simulations. The first is a long and laborious experimental technique, while the second is a long and laborious bioinformatic approach requiring high computational power.
To overcome these problems, a technique called Normal Mode Analysis (NMA) using the C-alpha representation of proteins was developed (Bahar et al., 1998). NMA exploits Hooke-type interactions between atoms and allows the generation of highly cooperative independent movements of varying amplitudes. Among these, there are long timescale movements of low energy cost that may represent the structural mechanism of a protein. The study of these movements allows us, among other things, to predict theoretical b-factors and generate different conformations.
The computational cost of this technique grows exponentially with the number of atom considered in the model. For this reason, most of the published models only use the alpha carbon of amino acid and ignore the impact of the nature of the amino acid. Despite this coarse-grained approach, NMA are able to predict b-factors to some extent compared to the crystal experimental data and can generate backbone conformational changes that link different known conformations of proteins.
With the goal of creating a more precise and realistic model, we have developed a new model using a coarse-grained approach using surface areas in contact between atoms (classified into different atom types) to represent the strength of interactions between alpha carbons. As a consequence the new NMA model is able to differentiate, in terms of molecular movements, between two otherwise identical conformations (at the level of C-alpha carbon atom positions) that differ in the identity of the actual amino acids present. For example, the new model can in principle differentiate between the wild type and a mutant of a protein in terms of their dynamic properties. A Monte Carlo approach allows us to obtain similar result to the best know NMA model on a similar database with respect to the prediction of experimental b-factors. We will test the new STeM method on two generated databases to predict b-factor change on ligand binding and on the impact of single point mutation.
Furthermore, we implemented a technique that allows us to generate an ensemble of conformations from combining different normal modes. These conformations are selected based on physical constraints (bond and angle stretching) allowing us to reproduce different known conformations of proteins using only the movement of the NMA model. We are implementing the NMA as part of our in-house genetic-algorithm-based docking software, FlexAID.
The use of the new method developed allows us to dynamic aspects of protein function without requiring long and laborious experimental or computational methods and properly accounts for the chemical properties of amino acids.
Bac. Pharmacology - 2010