About the method and the python package. Citation.
First MDMix methodology publication goal was to identify target binding sites and predict their druggability based on first principles. That is, to propose a measure of the ability of the protein (and in particular the pocket under study) to bind a drug-like molecule making use of a physics-based approach, namely Molecular Dynamics (MD).
It was demonstrated by multiple experimental studies (1,2) in the field of Structure Based Drug Design (SBDD), that target proteins have the ability to unspecifically bind small organic co-solvents in regions over the surface that correlate well with binding sites and important interactions for the drug binding affinity. Several computational methods emerged exploiting this idea of promiscuity to identify hotspots using small co-solvent molecules (e.g. isopropanol, acetonitrile, etc.). Many applications can make use of these hot spot information: they can be used to identify binding sites , predict its druggability , as pharmacophoric points for docking and virtual screening , for fragment growing , etc.
One of the most promising techniques for hot spot identification is based on Molecular Dynamics with co-solvent/water mixtures for solvating the system. Nevertheless, MD based methods are still the most costly approach in terms of computation time and user expertise and time needed for the preparation of the simulation and analysis of the results.
Computer scientist are in charge of overcoming the former limitation of speeding simulation calculations and, in fact in the recent years, the actual time-cost has been reduced enormously since GPU hardware was introduced in high computing facilities. It is expected that the speed of calculation will continue increasing in coming years, making this first limitation a matter of years to be solved. The second limitation will still be present and will become more problematic as the quantity of data generated will also increase making it more difficult to be processed and analyzed.
pyMDMix was developed to overcome this second limitation. It is written in python, designed as a python module and distributed as open-source under the GPLv.3 license to encourage developers interested in the methodology to contribute and keep it alive and usable.
The software, written as a python module, is also accessible from a command line executable program (the pyMDMix User Interface) that gives direct access to the standard set up and analysis procedures; therefor users will not need programming skills to employ the methodology.
Please, feel free to submit comments, report bugs or requests for improvements. Also leave some message if it's being useful! All feedback is really valuable!
MDMix related publications:
- Molecular Simulations with Solvent Competition Quantify Water Displaceability and Provide Accurate Interaction Maps of Protein Binding Sites. Daniel Alvarez-Garcia and Xavier Barril. J Med Chem. 2014.
- Relationship between Protein Flexibility and Binding: Lessons for Structure-Based Drug Design. Daniel Alvarez-Garcia and Xavier Barril. J. Chemical Theory and Computation. 2014.
- Binding Site Detection and Druggability Index from First Principles. Jesus Seco, F. Javier Luque and Xavier Barril. J Med Chem. 2009.
To cite the software, please refer to Reference 1 (D.Alvarez-Garcia and X.Barril, J. Med Chem. 2014).
Some of the applications are highlighted in the following sections as an example.
High occupancy regions of the surface identify high affinity interaction spots: The Hot spots
A simulation of HIV-1 Protease within an ethanol / water mixture (20% concentration) can be easily setup with pyMDMix. You won't have to worry about the solvent at all. You will only need previous knowledge on system preparation using Ambertools software.
Two energy grids are obtained after analysis of the simulation. The red surface corresponds to the hydroxyl head of the ethanol molecule. The green one corresponds to the methyl tail. Thus, red highlights polar interaction hotspots and green the hydrophobic regions with high propensity to interact with organic molecules.
When we superpose crystalized ligands, many hotspots overlay with described key interactions.
High density of hotspots highlight binding sites
Taking a quick look at the lower values of the energy grids obtained, quickly some regions of the protein are noticeable. They usually correspond to described binding sites (for small molecules and/or protein-protein interfaces).
A superposition of crystalized ligands is shown in the right image over the HIV protease structure.
Conversion to binding free energies is done by comparing the observed distribution with the expected one through the boltzman relationship. For this energy to be correct, the protein-solvent conformational space has to be correctly sampled.
Energy values are estimations of the free energy that interaction might provide to our ligand affinity. As we are not usually working at 1M concentrations, free energy values are analytically corrected to obtain Standard State Free Energies.
The clustered hot spots and its associated energy can be used to estimate a total binding affinity a pocket could provide if a perfect match with a ligand was ever met.
One of the main differences and advantages between using simple Molecular Interaction Potentials and the MDMix energy maps can be easily demonstrated when superposing the hydroxyl interaction probe and the water probe from MIP programs and from MDMix methodology.
The image at the right shows in yellow mesh the MIP map for hydroxyl probe (A) (which is almost identical to the water probe - data not shown-), in red surface the hydroxyl probe from the ethanol molecule from MDMix simulation (B) and in orange surface, the water map also from a MDMix simulation (C).
Clearly, (A) overlays with the water map (C) from MDMix, and the ethanol hydroxyl (B) is the only map matching the ligand's polar interacting moiety.
Water maps also give information
Clearly the explicit solvent simulations give us more information on the water hydration sites. As shown before, a water molecule was stuck in the cavity and the ethanol hydroxyl was forced to interact with it.