I have computed the binding energy of my protein to the ligand, initially by docking and after MD by MMPBSA. How do I relate these binding energy values to experimental IC50 that is reported in papers?
ideally the binding free energy i.e. MMGB/PBSA should follow the same trend as IC50. if you convert the binding free energy to IC50 using DeltaG = RTlnIC50 it should be almost similar to the IC50 but in my experience is not the case all the times. so binding free energy i.e MMGB/PBSA are good to show trend and compare two ligands in same receptor. Out of MMPBSA and MMGBSA I feel more comfortable to use MMGBSA as it faster with almost comparable results to MMPBSA. you can check this paper for reference http://pubs.acs.org/doi/abs/10.1021/ci100275a
You cannot directly transform IC50 into binding free energy; additional transformations are required. See, for instance http://en.wikipedia.org/wiki/IC50.
@Nitin: In my case, I cannot perform mmpb/gbsa as I am using the charmm forcefield and hence cannot use the AMBER for performing this calculations.
@Justin: From the link, I understand that relationship can be given by the Cheng-Prussof relationship. Unfortunately, I do not have the values for substrate concentration or Km from literature for these inhibitors. In that case, will it be sufficient if I show a similar relative trend with my MMPBSA rather than trying to match the values numerically?
MM/PBSA calculations are not restricted to any one particular force field or software. One can perform such calculations on the output of any of the major simulation software packages.
If you're trying to plot dG of binding from MM/PBSA vs. IC50, I'd say it's a bit suspect because, as stated above, there is not necessarily a relationship between the two, so it may not be terribly informative.
@Justin : Even though the MMPBSA analysis does not depend on the forcefield, the reason I could not go for the conventional AMBER MMPB/GBSA was that I could not find the parameters for the lipid substrates present in my protein while CHARMM36 had the parameters.So I resorted to gromacs in which I imported the charmm36 and used the utility g_mmpbsa ( a user contribution) and I am currently running the analysis using the same. Hope the methodology that I used, bypasses force field- dependent discrepancies.
But based your last reply, is there no way I could correlate my MD results to experimental values? (I thought that would be how we can actually validate the correctness of our simulations.). In this situation, what values do I choose as reference to I validate the deltaG binding values that I get from MMPBSA?
The only data to which you can compare computationally derived deltaG values is either a binding free energy from experiment (e.g. ITC) or Ki values from inhibition assays. IC50 values are more a function of solution conditions than anything else, so it's a rather weak argument to try to compare IC50 and deltaG.
@justin though we can't get IC50 value from deltaG BUT am i right in saying that deltaG of ligands should follow same trend as IC50 or there is no correlation at all
Possibly, but it's not guaranteed. IC50 is not a measure of affinity; deltaG is. I don't see what meaningful comparison would be drawn. Indeed there may be some coincidental trend that can be observed, but as stated above, it is a weak comparison, at best.
@Justin: Thank you so much for the answer. In my case, the inhibitor is clearly non-competitive for one substrate ( where the inhibitor binding did NOT affect the rate of binding of the substrate and was completely reversed upon incubating with substrate) and indirectly non-competitive for the other substrate ( where the rate of binding of the substrate is reduced with the inhibitor binding but was again reversed upon incubating with the substrate).
Thus in my case, Cheng-Prussof relationship might not be applicable. Our lab is entirely computational and from the literature, I have nothing but the IC50 values.
And I also see that MMPBSA binding energies are not falling in trend, which is expected.
Is there anything that you would like to suggest at this juncture?
Thank you so much for taking time in clarifying my doubts.
If the hypothesis was the MM/PBSA and IC50 are in some way related, it sounds like you've reached the expected conclusion - they aren't. Beyond that, there are too many things that I don't know, so I can't begin to suggest anything useful. Surely your supervisor and/or colleagues will have better insight than random people on the Internet :)
I do not have full text access to the same but the last line of the abstract says "IC50=Ki for non-competitive and uncompetitive and Cheng-Prussof is valid only for the competitive inhibitors.
Kindly provide your valuable suggestions in this regard,
Allostery is complicated. I think it is odd to ever equate IC50 and Ki, and I would venture that such an assumption only holds true in the simplest of cases. You already said that you plotted the results you got against IC50 and there was no trend. I would strongly suspect that, as with most allosteric processes, there is a complex network of intra-protein movements that affect ligand binding after binding of the inhibitor, i.e. different collective motions, occlusion of the ligand-binding site, etc. The affinity of the inhibitor for the allosteric site is not necessarily an indicator of how effective of an inhibitor it is; rather, the efficacy of imposing structural change on the receptor is what determines its ability to inhibit ligand binding. As to the specifics of how you should proceed, again there's not nearly enough information for anyone here to say.