I agree with Drs. Kamberaj and Matthews statements. In general, I have found the force-fields that govern the binding energies are not entirely reliable in docking programs that use them as part of the scoring function. My understanding of the literature is that many researchers put more value onto the resulting poses over the binding energy values. There are several papers that show a docking-->MD-->MM(GB/PB)SA workflow yields favorable results in terms of ranking ligands.
Experimentally speaking, binding energies are a good metrics for ranking ligands that target a specific pocket. However, due to the approximations made in theoretical modeling, one has to assess the theory level that is used in predicting the binding energies. Rigid docking by itself has many assumption built into it concerning the idea of modeling a ligand bound to a pocket, including the use of (in general) a solid-phase protein structure, the lack of ligand and binding site conformational dynamics, choice of scoring function, how water is modeled, etc. Never-the-less, many researchers have success in using docking for identifying lead compounds.
There are a number of articles out there that dive into your question and this topic, some of which are listed below.
Suggested reading:
1. Guimarrães, C. R. W. & Cardozo, M. MM-GB/SA rescoring of docking poses in structure-based lead optimization. J. Chem. Inf. Model., , 2008, 48, 958-970.
2. Guimarães, C. R. W. MM-GB/SA Rescoring of Docking Poses. Methods Mol. Biol., 2012, 819, 255-268
3. Alonso, H.; Bliznyuk, A. A. & Gready, J. E. Combining docking and molecular dynamic simulations in drug design. Med. Res. Rev., 2006, 26, 531-568
4. Bursulaya, B. D.; Totrov, M.; Abagyan, R. & Brooks, C. L. 3rd. Comparative study of several algorithms for flexible ligand docking. J. Comput. Aided Mol. Des., 2003, 17, 755-76.
5. Cheng, T.; Li, X.; Li, Y.; Liu, Z. & Wang, R. Comparative assessment of scoring functions on a diverse test set. J. Chem. Inf. Model., 2009, 49, 1079-1093
6. Englebienne, P. & Moitessier, N. Docking ligands into flexible and solvated macromolecules. 4. Are popular scoring functions accurate for this class of proteins? J. Chem. Inf. Model., 2009, 49, 1568-1580
7. Enyedy, I. J. & Egan, W. J. Can we use docking and scoring for hit-to-lead optimization? J. Comput. Aided Mol. Des., 2008, 22, 161-168
8. Ferrara, P.; Gohlke, H.; Price, D. J.; Klebe, G. & Brooks, C. L. Assessing scoring functions for protein-ligand interactions. J. Med. Chem., 2004, 47, 3032-3047
9. Foloppe, N. & Hubbard, R. Towards predictive ligand design with free-energy based computational methods? Curr. Med. Chem., 2006, 13, 3583-3608
10. Halperin, I.; Ma, B.; Wolfson, H. & Nussinov, R. Principles of docking: An overview of search algorithms and a guide to scoring functions. Proteins, 2002, 47, 409-443
11. Hou, T.; Wang, J.; Li, Y. & Wang, W. Assessing the performance of the molecular mechanics/Poisson Boltzmann surface area and molecular mechanics/generalized Born surface area methods. II. The accuracy of ranking poses generated from docking Journal of Computational Chemistry, 2011, 32, 866-877
12. Neudert, G. & Klebe, G. DSX: a knowledge-based scoring function for the assessment of protein-ligand complexes. J. Chem. Inf. Model., 2011, 51, 2731-2745
The binding energy should be given more preference while docking as the complex should be stable.
Thanks for your reply. But it's not always the case, you may right, if the question is dealt superficially, but not when you go in depth of it, in terms if the binding pocket of your protein is specific.
Because some research papers published in very good journals as well as pharmaceuticals company fellows working on docking and its optimization prefer good pose rather than the binding energy. And practically i have seen these cases when binding energy of a particular ligand has shown a very good binding energy but its pose was not good.
Any ways once gain thanks for your reply.
Yes you are right at this point. I agree with you. Even in the case when experimentally you will get a certain binding energy and Ka and Kd, you have to select the particular ligand whose binding energy and inhibition constant are very close to the experimental data(Ki=kd).
Best Regards,
Do you know of a ligand that has be shown to bind to the same site you have docked to? If so it may be possible to repeat the docking with that ligand and compare binding energy to the other ligand(s) you have used. I think you've actually answered your own question; sometimes you choose the result with the best binding energy, sometimes you choose the best pose. I think it depends on what you are trying to achieve. At the end of the day in silico docking can only get you so far, nothing beats wet-lab experimental follow up.
In general, the docking programs miss one important term in binding energy of ligand to receptor, which is the entropic term. This term can be decomposed in rotational, translational, vibrational, and conformational entropies, where the first two and the last one may be significant.
I think, from the information that you get from docking software poses are binding modes, that is the relative orientation of ligands w.r.t. protein and their respective conformations. The binding energy (or scoring) is more the number of important interactions, such as intermoleular, h-bonds, hydrophobic, which can be used to rank the poses bases on the number of counted favourable interactions.
For a more quantitative way of comparing the stability of complex with experimental results, poses taken from docking software can be further studied using for example MD simulations and compute the binding energies, which will include all terms, entropic terms as well.
Thank you all for your important suggestion and time given to this discussion.
Dr. Hiqmet, I have a question, if I am not getting wrong, you want to say, that the poses obtained after the initial dockings like grid based in case of dock6 are relatively more important than the binding energy and one should only consider or give importance to the binding energy calculated after running MD simulation on the perfect poses obtained through the initial grid based docking.
I think, this is very interesting point. You could say that one can do MD simulations using the replica exchange method, then in this case you can replicate one of the poses obtained, for example, for Docking, and since you can simulate all range of temperatures, you can obtain from the simulations all the binding modes (i.e., poses) that you get from Docking, but now you have also the Boltzmann weights for each pose. Of course, you never sure how accurate are the force fields for high temperature.
If that is the case, then we can think of doing configuration replica exchange simulations, that is fixed temperature of each replica (let say 300 K), but starting configurations for each replica to be the different poses you get from docking, then at the end you average out over all replicas. (In this case is important to start with really different configurations, such as these you get from docking, so the poses of docking are important, since you expect that they are different, you have more freedom on sampling ligand in docking.)
I agree with Drs. Kamberaj and Matthews statements. In general, I have found the force-fields that govern the binding energies are not entirely reliable in docking programs that use them as part of the scoring function. My understanding of the literature is that many researchers put more value onto the resulting poses over the binding energy values. There are several papers that show a docking-->MD-->MM(GB/PB)SA workflow yields favorable results in terms of ranking ligands.
Experimentally speaking, binding energies are a good metrics for ranking ligands that target a specific pocket. However, due to the approximations made in theoretical modeling, one has to assess the theory level that is used in predicting the binding energies. Rigid docking by itself has many assumption built into it concerning the idea of modeling a ligand bound to a pocket, including the use of (in general) a solid-phase protein structure, the lack of ligand and binding site conformational dynamics, choice of scoring function, how water is modeled, etc. Never-the-less, many researchers have success in using docking for identifying lead compounds.
There are a number of articles out there that dive into your question and this topic, some of which are listed below.
Suggested reading:
1. Guimarrães, C. R. W. & Cardozo, M. MM-GB/SA rescoring of docking poses in structure-based lead optimization. J. Chem. Inf. Model., , 2008, 48, 958-970.
2. Guimarães, C. R. W. MM-GB/SA Rescoring of Docking Poses. Methods Mol. Biol., 2012, 819, 255-268
3. Alonso, H.; Bliznyuk, A. A. & Gready, J. E. Combining docking and molecular dynamic simulations in drug design. Med. Res. Rev., 2006, 26, 531-568
4. Bursulaya, B. D.; Totrov, M.; Abagyan, R. & Brooks, C. L. 3rd. Comparative study of several algorithms for flexible ligand docking. J. Comput. Aided Mol. Des., 2003, 17, 755-76.
5. Cheng, T.; Li, X.; Li, Y.; Liu, Z. & Wang, R. Comparative assessment of scoring functions on a diverse test set. J. Chem. Inf. Model., 2009, 49, 1079-1093
6. Englebienne, P. & Moitessier, N. Docking ligands into flexible and solvated macromolecules. 4. Are popular scoring functions accurate for this class of proteins? J. Chem. Inf. Model., 2009, 49, 1568-1580
7. Enyedy, I. J. & Egan, W. J. Can we use docking and scoring for hit-to-lead optimization? J. Comput. Aided Mol. Des., 2008, 22, 161-168
8. Ferrara, P.; Gohlke, H.; Price, D. J.; Klebe, G. & Brooks, C. L. Assessing scoring functions for protein-ligand interactions. J. Med. Chem., 2004, 47, 3032-3047
9. Foloppe, N. & Hubbard, R. Towards predictive ligand design with free-energy based computational methods? Curr. Med. Chem., 2006, 13, 3583-3608
10. Halperin, I.; Ma, B.; Wolfson, H. & Nussinov, R. Principles of docking: An overview of search algorithms and a guide to scoring functions. Proteins, 2002, 47, 409-443
11. Hou, T.; Wang, J.; Li, Y. & Wang, W. Assessing the performance of the molecular mechanics/Poisson Boltzmann surface area and molecular mechanics/generalized Born surface area methods. II. The accuracy of ranking poses generated from docking Journal of Computational Chemistry, 2011, 32, 866-877
12. Neudert, G. & Klebe, G. DSX: a knowledge-based scoring function for the assessment of protein-ligand complexes. J. Chem. Inf. Model., 2011, 51, 2731-2745
Thank you Dr Kirschner that gives me a good list of papers to add to my collection
There is a slightly more fundamental problem at work here; it is not, even in theory, possible to compute the binding affinity for a ligand from a single static pose of a protein-ligand complex. Binding energy is the difference in energy between the starting, unbound, state and the final, bound, state. Using a single bound pose, or even a collection of them, ignores the energetics of the unbound protein and the unbound ligand and therefore cannot provide a binding energy. It has been shown in a few cases that RELATIVE energies of related compounds can be be computed with sufficient accuracy to be useful, but the computation of binding affinity from a given protein-ligand complex simply cannot be achieved.
In work presented as part of the series of SAMPL meetings over the past 4 years, prospective prediction of binding affinity has proven consistently beyond the reach of both docking and MD methods, which casts the successful retrospective prediction of binding affinity into some question.
Considering another, much simpler problem, may help in understanding the magnitude of our task: if we cannot predict reliably the transfer energy of a small molecule from gas to water with chemical accuracy (which we cannot with force fields), how can we predict the transfer energy of a small molecule from water to the protein binding site using force fields? And that leaves aside how we'd even begin to address protein entropy, which is never considered in scoring functions.
In response to Prof. Kamberaj's comment on entropy:
"This term can be decomposed in rotational, translational, vibrational, and conformational entropies, where the first two and the last one may be significant."
The third entropy, vibrational, is arguably the most important. Understanding ligand vibrational entropy reflects well width, which is vital in understanding small molecule thermodynamics.
So, if you know how to tell reliably if a pose from DOCK6 is good, independent of the DOCK6 energy scoring, then use the good poses. "Binding energy" from DOCK6, or any other docking engine, is a very poor approximation to binding affinity.
Thank you Dr Kamberaj, Dr Kirschner and Dr Hawkins for your interesting answers. It really helps me in understanding the importance of good poses a lot.
Dr Hawkins, would you please suggest me some good articles or research papers related to your conclusion, so that I could support my point well.
And another thing that I want to know is, About reliability of Grid_score obtained after grid based docking using Dock6.
Please help me with your valuable suggestion.
scoring functions sometimes perform poorly, so I suggest good docking poses as the better criterion
But while going for HTS or Virtual screening there are several ligands who show good pose, then how one going to extract the best, I think for the curation of the best ligand after sort listing the good pose, one should go for MD and MM(GB/PB)SA
If you know the likely kinds of interactions that can occur between the ligands and the protein binding site, then use visual inspection to identify the putative binders by performing protein-ligand interaction studies
MD and MM(GB/PB) SA will work if you have good starting poses for the MD simulation. Otherwise you may generate tons of data that are meaningless.
one of protein and ligand docking binding energy is -3.91 and another is -3.74 which is best docking results
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics