I would like to have an estimate of the binding energy due to the non-bonded interactions between a ligand and a binding site on a protein. I used to get this data from the output of docking programs such as VINA, but now I need to have the binding energy from an already "positioned" ligand (such as the complexes that can be found in teh RCSB database).
Does someone know a simple straightforward software I could use (maybe PyMol plugin?) or a web server performing such kind of calculations?
Hi, you can do it in autodock. If you include the 'epdb' keyword in a dpf (and omit all keywords which invoke a search), autodock will score the molecule.
[You will need to create pdbqt files for each of your 'manually generated poses' and a separate dpf for each one.]
Here is a complete example of an epdb docking parameter file:
autodock_parameter_version 4.2 # used by autodock to validate parameter set
outlev 1 # diagnostic output level
intelec # calculate internal electrostatics
ligand_types C HD N NA OA # atoms types in ligand
fld rec.maps.fld # grid_data_file
map rec.C.map # atom-specific affinity map
map rec.HD.map # atom-specific affinity map
map rec.N.map # atom-specific affinity map
map rec.NA.map # atom-specific affinity map
map rec.OA.map # atom-specific affinity map
elecmap rec.e.map # electrostatics map
desolvmap rec.d.map # desolvation map
move lig.pdbqt # small molecule
about 17.6 22.2 32.6 # small molecule center
epdb # small molecule to be evaluated
Here is part of the corresponding output which also includes a detailed "Ligand Intramolecular Energy Analysis":
epdb: USER Estimated Free Energy of Binding = -6.54 kcal/mol [=(1)+(2)+(3)-(4)]
epdb: USER Estimated Inhibition Constant, Ki = 15.95 uM (micromolar) [Temperature = 298.15 K]
epdb: USER
epdb: USER (1) Final Intermolecular Energy = -7.14 kcal/mol
epdb: USER vdW + Hbond + desolv Energy = -6.33 kcal/mol
epdb: USER Electrostatic Energy = -0.81 kcal/mol
epdb: USER (2) Final Total Internal Energy = -0.20 kcal/mol
epdb: USER (3) Torsional Free Energy = +0.60 kcal/mol
epdb: USER (4) Unbound System's Energy [=(2)] = -0.20 kcal/mol
epdb: USER
epdb: USER
Most ligand docking programs will have options for re-scoring docked ligands which you could use, so just re-score the ligand from the pdb file.
But whatever you get from a simple scoring function will only be a crude estimate, the correlation between 'scores' and experimentally determined KDs is notoriously weak.
Calculating Free Energies of Binding is not a trivial problem as you attempt to calculate the (rather small) difference between two (rather big) quantities, and your errors come from calculating the latter ...
I believe you can use NNScore, a specialized software which may deal with your AutoDock (Vina) customized input files (PDBQT) to compute binding free energies. It is freeware. However, to make it very easy, DSXONLINE is a web server for this purpose (registration is required). If you want to graphically analyze the protein-ligand interactions (other than Pymol and Chimera), you should go for HYDE module available in LeadIT software (commercial, but trials are available). It graphically aids you to recognize the region of contribution and detrimental effects brought up by ligand binding with amino acids. It also generated Ligand Efficiency (LE) for cross-docking experiments.
Hi, you can do it in autodock. If you include the 'epdb' keyword in a dpf (and omit all keywords which invoke a search), autodock will score the molecule.
[You will need to create pdbqt files for each of your 'manually generated poses' and a separate dpf for each one.]
Here is a complete example of an epdb docking parameter file:
autodock_parameter_version 4.2 # used by autodock to validate parameter set
outlev 1 # diagnostic output level
intelec # calculate internal electrostatics
ligand_types C HD N NA OA # atoms types in ligand
fld rec.maps.fld # grid_data_file
map rec.C.map # atom-specific affinity map
map rec.HD.map # atom-specific affinity map
map rec.N.map # atom-specific affinity map
map rec.NA.map # atom-specific affinity map
map rec.OA.map # atom-specific affinity map
elecmap rec.e.map # electrostatics map
desolvmap rec.d.map # desolvation map
move lig.pdbqt # small molecule
about 17.6 22.2 32.6 # small molecule center
epdb # small molecule to be evaluated
Here is part of the corresponding output which also includes a detailed "Ligand Intramolecular Energy Analysis":
epdb: USER Estimated Free Energy of Binding = -6.54 kcal/mol [=(1)+(2)+(3)-(4)]
epdb: USER Estimated Inhibition Constant, Ki = 15.95 uM (micromolar) [Temperature = 298.15 K]
epdb: USER
epdb: USER (1) Final Intermolecular Energy = -7.14 kcal/mol
epdb: USER vdW + Hbond + desolv Energy = -6.33 kcal/mol
epdb: USER Electrostatic Energy = -0.81 kcal/mol
epdb: USER (2) Final Total Internal Energy = -0.20 kcal/mol
epdb: USER (3) Torsional Free Energy = +0.60 kcal/mol
epdb: USER (4) Unbound System's Energy [=(2)] = -0.20 kcal/mol
epdb: USER
epdb: USER
I don't know any web servers or freely accessible tools/plugins since usually I use a tool implementable on Sybyl (Tripos) called HINT (Hydropathic INTeraction; produced by edusoft)
You can also use database "PDB bind" to such information. http://www.pdbbind.org.cn
You may use PISA in CCP4 to calculate the interaction energy. Here is the web server: http://www.ebi.ac.uk/msd-srv/prot_int/cgi-bin/piserver. Good luck!
We did such things with the QMScore. This is included in the Suite of the guys from QuantumBio Inc. It could be fully integrated as a plugin into MOE and it uses semi-empirical methods. therefore it is more trustable, than any other thing I've seen before. I have to say it's not for free, but very fast and efficient in such questions. Also the visualization is straight forward. You could see the interaction terms for each amino acid that is involved in the binding of the ligand as a color coding. I recommend to take a look at their web-site.
I think Autodock will be best. you can also use web servers like Haddock
haddock.science.uu.nl/, PatchDock Server: An Automatic Server for Molecular Docking
bioinfo3d.cs.tau.ac.il/PatchDock/, Hex Protein Docking Server
hexserver.loria.fr/ and many more.....
To calculate the binding energy you should also include entropic term, which is related to rotation, translation and vibration components of entropy. A cheap way to do this is to use molecular dynamics, then from trajectory use a mmpb or mmgb method to decompose binding energy into all these components including interaction terms. Translation and rotation change of entropy during binding may be important for your case.
you are asking for a challenging task. in computational chemistry you can have different accuracy of binding energy calculations, depending on what your are looking for. fast and rapid and "relative" calculations you may want use MM/PBSA and or MM/GBSA approaches. In amber there is a tool called mmpbsa.py and in gromacs gmxapbs (unofficial tool, developped from our laboratory: //www.plosone.org/article/info:doi/10.1371/journal.pone.0046902). If you have a set of pdb files you can use dfire (http://sparks.informatics.iupui.edu/hzhou/dfire.html) and fastcontact (http://structure.pitt.edu/software/FastContact/) tools. Remember, these approaches are useful only if you compare a set of ligands and/or mutants of you receptors.
By definition such a thing cannot be done without knowing the unbound state of the ligand and the unbound state of the protein, along with the mutually bound state. There is no way that just scoring the protein-ligand complex can tell you, absolutely, what the binding energy is. You might be able to obtain relative rankings for closely related ligands (due to fortuitous cancellation of errors), but an absolute affinity will never be calculable just from knowledge of the bound state. Also, as mentioned above, scoring might tell you a good deal about the enthalpy of the interaction, but you obviously must have the entropic component too. This is accessible to some degree of accuracy for the ligand, but is elusive for the protein.
Now if what you actually want is the binding enthalpy then you might have more success.
You may also try pair potential calculations to check the enthalpy contribution is upto the level as predicted by the docking program. This will give you a rough estimate. Please perform MD to better understand the non-covalent interactions.
The following is applicable is what you're interested in is just a binding score from your program:
If you want to compare the your PDB ligand to other ligands, you should use the same forcefield / method, as scores from different methods should not be compared! So, if you previously generated your scores with VINA, you need to score your bound ligand with VINA. This might seem simplistic, but why don't you just re-dock your ligand with the program you normally use? After all the first validation that your docking program & parameters are appropriate is that they should reasonably reproduce the conformation of your reference ligand. And if they can't, why would you trust the scoring results at all?
Now, about the "real" binding energy, I could have made a copy-paste of Paul's comments as they make perfect sense.
i have seen some studies where MMGB/PBSA were calculated after minimization and no simulation was performed to refine docked poses. is this a correct approach?
in normal cases binding free energy should be average of multiple simulations. So how how much reliable is calculating binding free energy from a single pose? I have performed docking and selected poses from visualization as docking scores are not working for me at all
Hi Gianluca, may I suggest you taking a look the web application Kdeep, available free of charge at the PlayMolecule suite (www.playmolecule.org). Convolutional neural networks are used to predict binding affinity (Kd and binding free energy) for a set of docked protein-ligand complexes. Therefore, you can input your VINA-docked ligand structures in a sdf file and the PDB of the protein to get estimations for the binding free energy.
I would recommend AutoDock or AutoDock Vina as the "easy way" to estimate binding energies. As mentioned before, you should look at relative values to compare ligands. In general, anything below -8 kcal/mol is good. The other option is MM/GBSA from Schroedinger. It works great but the value is at best relative in a set of ligands. It gives very good predictions and we use it all the time.
Getting an absolute value for binding energy is a tough cookie. You are better off not attempting it as the results may not be as they are measured in the lab.
- Badges
- Science topic
- Similar topics
- Chemistry
- General Biochemistry
- Proteomics