GROMACS version: 2020.7
GROMACS modification: No
Here post your question
Hello everyone, is it possible to prepare a portein for molecular docking using MD simulation in NVT and NPT ensembles and then use the output file in one’s desired solvent and temperature as an input file for a molecular docking software such as Auto Dock? as you know .pdb files available on the Protein Data Bank are all produced in aqueous medium and near 25C.
I appreciate any help in advance.
What are the possible difficulties you’re thinking of, performing docking with a solvation term corresponding to a non-aqueous environment? It can be highly non-trivial as some empirical scoring terms are adjusted to the special properties of water. On the other hand, some people use water admixed with non-polar fragments to encourage the exploration of cryptic pockets, something that can be then used for docking in aqueous solvent.
Also note that few protein structures in the PDB are actually produced/solved at 25ºC, but again, if your goal is to explore cryptic pockets, you anyway want to have an ensemble view, perhaps even biased towards pocket opening.
I really appreciate your response.
The thing that really confuses me is that I have used an enzyme as a biocatalyst in acetonitrile as solvent and I want to use MD as a way to explain the stereoselectivity that was observed but since docking and MD softwares are designed for aq. medium I don’t really know how to prepare my system(including protein and protein-ligand complexes) in the most reliable way.
What would you recommend?
Thank you
What is the state of knowledge there, do you know the binding pose of the ligand in water, or are you using the MD simulations to figure that out? MD codes aren’t specifically designed for aqueous medium (or are very weakly biased towards aqueous solutions), so as long as you can rely on statistics from equilibrium or enhanced-sampling MD (or free energy methods), I would say it’s way more reliable. I’d say you can still use docking to get “good” starting positions for simulations (as long as these simulations are long enough), but not necessarily trust the scores that AutoDock will produce.
Thank you for your informative response.
Regarding your question I must note that the reaction that I’ve worked on actually don’t proceed in water at all!
Is it necessary to perform MD for a long time? The reason I’m asking this question is that one of the components in our reaction ,despite having better binding energy (according to auto dock studies), moves completey out of enzyme’s active pocket during the first one or two nano seconds of the simulation resultion in RMSD values as large as 10 or higher! (which actually is a good thing since we have hypothesized that one part of the reaction is occuring in the solution rather than in the active site). with all that being said is it enough to perform MD for 10 or 15 ns or should I go for more than 80ns?
I really appreciate your help.
It’s really a question of the purpose of your calculation. If you want to validate a single binding pose and the ligand unbinds rapidly, then it probably isn’t a stable pose; but I myself remember cases where the ligand would just shift to a nearby position and remain stable there, so watching the trajectory (rather than just tracing RMSD) helps.
One approach one might follow to identify binding pockets would be to add multiple copies of the ligand (assuming it’s reasonably small) at random positions and run several long simulations to find out frequent interaction sites - then we’re talking at least microseconds of total simulation time to get trustworthy insights, although this will depend on the sizes of the ligand and the protein.
In the approach you’re pursuing - using seeds from docked ligand geometries - I would say that 100s of nanoseconds is the minimum residence time to consider something stably bound, but also look out for possible nearby poses, i.e. situations like those described above: your ligand might move a few Anstroms and find a more suitable position within the same extended cavity or groove.
Note that single 10-50 ns simulations were considered acceptable demonstration of stability a decade ago, but today we’re in an era where a single decent GPU might get you up to 1000 ns/day on reasonably small systems, so the expectations have gone up accordingly :)
Thank you so much for your answers.
Best of regards.