Protein/Molecular Docking

Here there are some summaries (by the moment one) of some papers that I have read. I hope it is useful for anybody that want to work on that to help the medical society.
Protein docking is a very important, and interesting, challenge. It is known that the 3D structure of protein-protein complexes helps to understand and determine the function of molecular systems. However, it is not easy to figure out which is best way two protein interacts to form a molecular complex. There are several ways to deal with this problem.


Authors: Henry A. Gabb, Richard M. Jackson and Michael J.E Sternberg
Published at: JMB (1997) 272, 106-120
Title: Modelling Protein Docking using Shape Complementary, Electrostatics and Biochemical Information
My Summary and comments:
Authors propose a method based on the Katchalski-Katzir et al. proposal. However, this method improve the quality of the results thanks to some additional filtering that they add, and the use of higher resolution on the grids. Katchalski-Katzir et al. proposed a protein docking algorithm based only on shape complementarity.  Shape complementary process is done taking advantage of the fast Fourier transform (FFT) and Fourier correlation. In order to do that, proteins A and B are discretized in 3D grids, and different values are assigned to each node of the grid. The values assigned to each 3D grid node for protein A are different depending on if the node is at:
  1. Open space (value 1)
  2. Core of molecule (value d)
  3. Surface of molecule (value 0)
For the molecula B there are two possible values: 1 for inside of the molecule, 0 for open space. With that assignation, and using the Fourier theory, one can compute which is the shape complementary of the two proteins "very fast".

Herny A. Gabb et al. add different filtering processes to this shape complementary process. They add and analyze two different kind of filtering: one based on electrostatic propierties of  the proteins and another based on the biological information available for the proteins we are dealing with.
  1. Electrostatic complementary: Electrostatic interactions help to distinguish true and false positive dockings. This filter is based on a simple Coulombic model. They build two electrostatic grids for each protein to be Fourier transformed and correlated. They use the electrostatic correlation score as a binary filter at the end of the Shape Complementary process.
  2. Based on  available biochemical information:  Knowledge of the location of the binding site on one, or both proteins drastically reduces the number of possible allowed conformations. They apply different filtering degrees depending on this information (loose, medium, tight).
Finally, local refinament can be applied to the most reasonable predictions. During refinement each geometry is shifted (in each direction) and rotated (in each direction) slightly to find the highest surface correlation score in the local space. Thinner surface thickness has to be used to improve the results. If we use the same surface thickness than for global search, refinament can not improve the results as evaluated by the rank of the first good solution (however, refinement tends to improve final RMS).

In this paper, they perform a protein study for six enzyme/inhibitor and four antibody/antigen. As other researches comment, antibody/antigen problem is more difficult than enzyme/inhibitor.


Please, if you know any paper that you think is interesting to read, let me know to my e-mail address (djimenez at somewhere in, you know... don't like too much s-p-a-m) .