I have a compound, which we expect interacts with water. Experimentally the level of hydration in a tissue samples changes upon including this compound. We aren't exactly sure, though, how (if) water molecules are non-covalently interacting with the compound. We have identified four different sites which are very polar in nature. I wish to get the relative energy (relative to the unbound state of the compound) of water binding using 8, 10, 12, and 14 explicit water molecules using Gaussian. I have based these number on a number of studies on the hydration of aldehydes.
I have a good geometry of the compound (optimised separate). I want to:
a) identify the key water binding group (so a separate relative energy calculation per potential binding group),
b) calculate the relative energy of a totally immersed compound i.e., explicit water molecules around the 4 potential sites (obviously depending on (a)).
I am struggling though to justify (I have none) the position of the water molecules around each key site in my initial geometry. What would be the best technique? I have tried an energy minimisation using a Universal ForceField (UFF), but depending on the water molecule orientation this can separate molecules well over 2.5 Angs apart.
Is this a case where I should try constructing the possible hydrogen bond network (water-water-compoundgroup) by hand?
Finally, would it make more sense to model this in a vacuum, rather than an accompanying implicit water model? I am concerned that an implicit water molecule could interfere with intermolecular hydrogen bonds between water-water and water-compound.
Many thanks
Anthony
I'd recommend using classical MD first: you can run a short simulation (.5-1 ns) to allow the water network around your solute to relax. Then from the trajectory you can extract several frames with a given number of closest solvent molecules, and use them as initial structures.
This is not a trivial way to do this, but in my opinion the least arbitrary one -- you get initial frames sampled according to the statistical distribution. In fact, it is a quite common practice to generate initial structures for QM or QM/MM calculations from much longer (but much less exact) MD simulations in order to achieve good conformational sampling.
I'd recommend using classical MD first: you can run a short simulation (.5-1 ns) to allow the water network around your solute to relax. Then from the trajectory you can extract several frames with a given number of closest solvent molecules, and use them as initial structures.
This is not a trivial way to do this, but in my opinion the least arbitrary one -- you get initial frames sampled according to the statistical distribution. In fact, it is a quite common practice to generate initial structures for QM or QM/MM calculations from much longer (but much less exact) MD simulations in order to achieve good conformational sampling.
You can used MD simulation, may be Amber and then you can analyzed the trayectories using the keyword hbond of cpptraj in order to know the ocupancy of every hydrogen bond with a particular group of your compound
A suggestion from an experimentalist: you may evaluate the relative hydrogen bond donor (or acceptor) strength of potential basic sites from empirical scales (enthalpy or Gibbs energy scales). For example:
Basicity scales: J. Med. Chem., 2009, 52 (14), pp 4073–4086; J. Org. Chem., 2010, 75 (12), pp 4105–4123
Acidity scale: J. Phys. Chem. B 2014, 118, 4605−4614
Jean-François Gal, thanks for your reply. I always appreciate any insight from an experimentalist. I will look over those references.
Many thanks
Anthony
Welcome! If you need more details, we can discuss about the different scales devised for H bond (my colleague Christian Laurence et al. measured thousands of interaction energies).
J-F
I had done some calculation using 1 and 4 explicit waters in implicit solvation for nucleobases. As other people have suggested, one way find the appropriate site for the explicit waters is to run a short MD run (which is not very easy). You can check this paper:
dx.doi.org/10.1021/jp5088866
Dear Anthony,
I agree with others in that it is good to have an MD calculation as a starting point and use between 5 to 10 systems as your initial point. I believe adding the implicit water (continuum models) is essential. From my experience in simulating NMR and EPR parameters and also trying to correlate the orbital energies with some observables, it is normally necessary to add both implicit and explicit models simultaneously.
Dear Anthony,
Did you consider the Effective Fragment Potential (EFP) approach implemented in GAMESS-US program? It was thoroughly elaborated to model water molecules to study aqueous solvation effects, at the RHF/DZP or DFT/DZP (specifically, B3LYP) levels and could be combined with a continuum PCM model.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools