Very often water molecules are taking part in the hydrogen bond network established between protein and ligand. They may be important for binding affinity and specificity.

Thanks for the comment Adrian....But still the questions are remaining. Even if we cannot locate all the water molecules around the protein there is no meaning that protein structure contain only few water molecules. As the Mathews coefficient suggests where is the other water molecules. Then how we can say that these conserved water are only important playing an important roles? 

The thermodynamics and kinetics of long-residence water molecules interacting with a protein are different of that of other water molecules. The conformational flexibility and transitions can be influenced as a result. Please look at our paper. It should be useful to you. (http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1002394) 

As Abdallah said, long-lived water molecules within or close to the binding interface are important, as they are involved in the hydrogen bond network. In addition, trapped water molecules (not exposed to the bulk solvent, or mostly buried) inside the macromolecule structure, whether or not they are in the binding interface, are also key structural and functional elements.

You can take a look to the papers from Irene Luque's group at the University of Granada. They have carefully characterized the contribution of water molecules to the thermodynamics of ligand binding.

Dear Ahmad and Adrian thanks for the reply. I have already seen the plos comp biol paper. It's very interesting. What I have done so far is, with a modified water box I have done simulations for 50 ns. The simulations were done in triplicate to avoid some random error or bias. I have deleted some TIP3P water molecules from 3A radius from the Crystal water molecules and made a holo space. Then I performed MD and found water molecules are coming to the original position where the crystal water molecule was there. Then these water molecules are staying there for a long time. I calculated the occupancy of this water molecules in all the snap shots and in 95% of the snap shots these water molecules are staying. Then I performed in vacuum MD simulations (in triplicate) to see the effect in the absence of water molecules. Then I calculated the chi angle and fluctuations of active site residues from both simulations. Also measured the geometry of the pocket. I found the MD simulations performed in the absence of water (in vacuum) was because of the deletion of whole water molecules from the protein and not because deletion of conserved water molecules only. How can l address this? Whether this in vacuum simulations give any insights? Or by just demonstrating long resident water molecules are there it self is a good result? 

I do not think vacuum simulations would answer the role of the long resident water molecules in your protein. You need to eliminate the long-residence waters alone. You can do this for example by introducing a mutation or creating an identical residue with a modified atom type at the interaction site that hate water (use NBFIX, look for example at http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1004469). Alternatively, patching the interaction sites of these waters with a dummy atom that repels water. Another thing that can be useful is to manually eliminate the charge from the psf file at the interaction site of the long-residence water molecule.

Dear Ahmad. That's a good idea.  As you suggested mutation studies will give a clear idea. I will do that. I will update you with the results. Thanks for the suggestions. 

Dear Ahmad. Can you please explain me "Alternatively, patching the interaction sites of these waters with a dummy atom that repels water."

How many conserved water molecules are located in the binding site? Is the binding site accessible to bulk water?

Have you considered MD simulations with the conserved water molecules carrying zero partial atomic charges? (This can give you some information about the electrostatic pre-organization of the binding site for the specific arrangement of the water molecules.)

If the number of conserved water molecules is small, one can try a step-wise Free Energy Perturbation (FEP) simulation of the annihilation of the conserved water molecules: 1) gradually decrease the charges to zero; 2) gradually decrease the van der Waals radii to zero.

The Linear Interaction Energy (LIE) method could also be used: the active site amino acid residues can be treated as "ligands." (This could be a useful way to quantitatively compare different protein variants.)

---

All this assumes that there is no water exchange between the binding site and bulk water.

What Martin is suggesting will not work as the modified water molecles will be replaced by other bulk water molecules. @Dileep, I think the best is that you introduce a new amino acid that is similar to the one at the interaction site of the long residenence water. The new amino acid will differ from the original by one  atom at the interaction site. You just label the name of this atom differently in the parameter and topology files and keep all of its charge, vdw and bonded parameters the same. You then augument the parameter file with an nbfix term to repel waters from the interaction site. The practically will keep all other interactions the same except for water at the specified site.

Dear Martin I have only four conserved waters. I have already calculated the normalized b factor, SZmap and prime energy for all four water molecules and based on that a heat map was generated. Also calculated the overall energy contributed by these water molecules to the total energy. It was calculated through foldX. 

Dear Ahmad. Theoretically, we have to see the motion of loop in the absence of water molecules. So even these dummy atoms or charge modifications will not give the Right answer because the atom which present there will hinder the movement. So how we can address this case? That was the reason I decided to delete water molecules. Anyway I will try that too

Dear Dileep,

You will not hinder the movement of the loop since you are only modifying long-residence water interaction-site coupling to water molecules. This can possibly be done through NBFIX directly (depending if the atom type at the interaction site is unique in your protein) or by introducing a new atom type that replaces the interacting atom while keeping all nonbonded parameters and bonded parameters identical to the original one, then use NBFIX to modify its vdw interaction with water alone  to make it repulsive(see also https://www.ncbi.nlm.nih.gov/pubmed/28252958) .

If you decided to go to the dummy atom route, the dummy atom will have zero vdw radius and zero charge. It will effectively have only Lennard-Jones repulsive part with water by assigning suitable parameter values between water oxygen and the dummy atom through NBFIX charmm force field (see for example http://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1000435)

Dear Ahmad. Thanks for the detailed description. So far I have done the simulations with Desmond which will not allow to modify or incorporate any dummy atoms. I have to start either with gromacs or Amber to address the problems. Thanks for your valuable suggestions. I will try to do that. 

It is possible to define dummy atoms in Desmond.

---

Not a Desmond user.

Dear Martin. It is possible to define dummy atoms using Schrodinger

 But it will not take for the simulation. I think so. If you know how to do it, please suggest me some tips

Desmond can be used without the Schrodinger suite of programs.

Desmond can be extended by user-supplied plugins.

Desmond can be used to run free energy perturbation simulations, e.g. by changing a "real" atom into a "dummy" atom; if it can do it in the "forward" direction, it can do it in the "backward" direction as well.

---

Not a "Schrodinger" user.