'molecule-1' binds to a pocket in a given protein, while 'molecule-2' does not bind. However, using MD, both molecules fit well in the active site (pocket). Why? I can only suggest that the difference is not in binding (adsorption) but in diffusion: it may well be the 'molecule-2' does not diffuse as well as 'molecule-1'. How can I demonstrate using MD that 'molecule-1' binds whilst 'molecule-2' does not bind? Is it 'only' a matter of exploring diffusion through the accessible channel? What time scale (how many nanoseconds) should I need to see different diffusivities? Any other suggestions? I'm using LAMMPS software: perhaps better to use AutoDock Vina or other?
In order to get a more quantitative order of the affinity of your ligands to the protein, you can combine MD simulations with MM-GBSA or MMPBSA methods for calculations of binding free energies. Then, you can decompose it into electrostatic and other contributions, including entropic ones.
Alternatively, in order to have an idea of long range interactions and specific interactions, you can also permform steered MD simulations, non equilibrium, to get the PMF, which is basically pulling apart ligand and protein using external forces of different magnitude.
Yes, I use accurate electrostatics and a flexible model. Forcefield is UFF, and atomic charges have been obtained from a quantum semiempirical calculation on the protein using PM7/MMOK within MOPAC-2013 software.
Hi,
I do not know the molecules that you are trying to docking or docking program used. You're doing the same simultaneous docking protein? I do not understand why you say that using DM molecules fit well on the site since DM is not docking. Thus I suppose that the DM was taken after docking which allowed a greater number of conformations of ligands and hence better fit.
It is unlikely that semi-empirical methods can produce accurate electrostatics.
I would use at least MP2/aug-cc-pVDZ
How the water does take into account?
In order to get a more quantitative order of the affinity of your ligands to the protein, you can combine MD simulations with MM-GBSA or MMPBSA methods for calculations of binding free energies. Then, you can decompose it into electrostatic and other contributions, including entropic ones.
Alternatively, in order to have an idea of long range interactions and specific interactions, you can also permform steered MD simulations, non equilibrium, to get the PMF, which is basically pulling apart ligand and protein using external forces of different magnitude.
I agree with Dr. Kamberaj. Ultimately, the strength of a ligand-target interaction will come down to the relative binding free energy. There is numerous research articles that have used the methods he mentioned to investigate the questions that you are asking.
As for the electrostatics discussion, I would not be so absolute about saying one QM method over another provides you with better partial atomic charges. As Alexey Nikitin mentioned, MP2 theory with a larger basis set is probably more reliable at generating molecular electrostatic potentials (MEPs) than semi-empirical theories. However, I would do some literature search to see if this is indeed true or not. But we are talking about partial atomic charges, which are not an experimental observable and are not derived directly from the wavefunction as MEPs are. It is also believed within the AMBER community that partial atomic charges dervied from MEPs require some dampening in order to approximate solution-phase dipole moments (i.e. resp-derived charges). Ultimately, your scheme for deriving partial atomic charges require some validation, rather than assuming they are correct.
I am also not convinced that diffusion rate of the molecule will account for the differences in binding. I think this is a thermodynamic issue and not a kinetic issue, if both molecules reach the same binding pocket.
One last thing - both molecules may stay bound on the time scale of your simulation, meaning that it is possible that you haven't ran your simulations long enough to observe the dissociation event.
Best regards,
K
Thanks to everybody for the answers so far. A few comments ...
To Elias: yes you are right, the initial conformations of both systems (molecule-1/protein and molecule-2/protein) are supposed to be close to the relative minimum enthalpy conformation.
To Alexei: the protein contains 12000 atoms. MP2 does not seem the best option. I agree with the electrostatic comments by Karl. To start with, I did not considered water, but thanks for the suggestion: I agree this could be a determining factor.
Finally, and according to the answers so far, I will try to find the minimum energy starting from selected MD conformations in both cases (molecule1/protein and molecule2/protein).
MP2 for the ligand only. For protein there are charges of the usual force fields.
Water should be considered definitely, аt least with the simplest GB model, for example Sheffield model.
Try to perform a charge analysis of the ES complex for both the substrates and see the differences in values between the binding site and each substrate.
An idea should be a DFT CLUSTER approach.
To have some kind of PK/PD issues relating to binding and non-binding molecules, then I am getting curious to know how different are the two molecules chemically ?
In addition, If all the parameters used for simulation are supposed to be correct then mol-2 would dissociate but the time scale at which this would occur, i think, cannot be estimated or known, you might have to continue you simulation for a while.
I agree with Dr.Karl that the time scale may be an issue. I have worked for a similar problem and the net outcome was that due to the complicated energy landscape and the statistical nature of the simulation, one should not rely too much on seeing a dissociation event happens without intervention. Also, i do not know why you use the UFF, its use is very restricted, if any, for proteins and you must use other more advanced FF such as AMBERFF99SB-NMR or similar. I do not think there is a problem with the partial charges, first make sure to remove the bias from any other source of error then think about partial charges. Also, you might consider running an SMD or adaptive biasing force to trace the energy profile in dissociation, perhaps there is difference in the barriers that can cause one of your molecule to bind stronger than the other. I do not think that either MMGBSA or MMPBSA will help in your case.
- Badges
- Science topic