I don't think anyone can give you a specific time for this. it depends on the convergence of the simulation and the results you're looking for. If I were you, I  will run for 200ns or more with a time step of 10ps and see how much I can conclude if it is not sufficient I may extend the simulation or run for multiple conformations.  

Dear Nikhil Maroli, I appreciate your answer but the query is related to the criteria for the selection of poses which can be employed for the pharmacophore generation. The problem is which and how many poses to select for the pharmacophore construction. Though there are numerous papers published in this domain but none of them have clearly mentioned the criteria for the selection of the poses.

Dear Manhas

The approach sounds interesting. I think it's very difficult to generalize the appropriate time to extract the ligand or protein poses out from the simulation trajectory. At this point, I want you to remind that it's the features, conformation and interaction pattern which someone is looking not particularly the time. Also, prior to selection of time interval, the time scale of the biological process also needs to be considered. 

Other than that, I will suggest you to look into the literature and try to find out that what others have done.

Dear Anu, it depends upon your requirement to extract conformations from the trajectory suitable to develop MD-based pharmacophore modelling. Rather generalising the concept, I would ask you to search MD related papers for your protein-ligand complex. If someone from the literature had shown the changes in the global (protein) as well as ligand conformation changes (to estimate binding affinity), you may need to simulate your complex with the same simulation time as well as analyse the dynamic snapshots (i.e time interval where protein-ligand complex was exported) so that you can able to reproduce the known "MD observations and consequent conformational changes". However, the snapshots you retrieve from the trajectory at equal time intervals must be subjected to short energy minimization to reduce any bond and geometry inconsistencies (i.e this conformation may correspond to relative higher energy compared to your initial state). Again, depending upon your inputs (i.e. energy-minimized protein-ligand snapshots), you can develop both (i) ligand and (ii) receptor-based pharmacophores. For case (i), you can only retrieve the ligand conformation from each of the snapshots by deleting the protein chains and developing a model. If you switch-off the conformers generation step during pharmacophore generation, you may create ligand-based model which only show the hypothesis containing the pharmacophore features which did not deviate from the initial positions (say scaffold or ring systems) and can help you to show the most fluctuating (not represented by hypothesis due to distance tolerance) and static groups in the model. I think that this way of ligand-based models serves no purpose for further design or optimisation. The second case (ii) will help you to recognise both pocket lining residues which formed non-covalent contacts with the ligand functional groups and its participating ligand functional groups in such interactions. Moreover, you can quantify the proportion of amino acids involved in establishing such contacts throughout the trajectory and the essential features of the ligand. No matter, you have to validate the resultant models before you attempt to screen chemical libraries.

 Thank you, Prasanth Sir and Mohd. Athar..

I have one question related to receptor-based pharmacophore.

Is it possible to generated the receptor-based pharmacophore using Discovery studio?

Yes, you can generate it using Discovery Studio.

Even, good protocol (panel) and tutorials are provided with the DS suite.

for more details, you can check the attached file.

Thank's Mohd. Athar

Structure-based pharmacophore modeling is of two types: Macromolecule based (only protein) and Macromolecule-Ligand based modeling (protein-ligand).

It was helpful but I was asking about the macromolecule-based pharmacophore modeling. Can we generate it using Discovery Studio or in other words, can we select the amino acids manually for the generation of the model or if there is any other tool. 

Dear Anu,

Yeah, it is possible. Please go through this article: Steindl T, Langer T. Influenza virus neuraminidase inhibitors: generation and comparison of structure-based and common feature pharmacophore hypotheses and their application in virtual screening. J Chem Inf Model. 2004;44(5):1849–1856.

You need to know the time it takes to comprehensively sample all conformations of the binding site that would be "seen" as different by the ligand. This time must be comparable with the time it takes in the real world. Based on what we know about the binding kinetics and thermodynamics, these times should be in the range of microseconds to milliseconds. Since those times are usually hard to achieve, the rule of thumb would be: the simulation time should be as long as your institutional computing resources and time constraints of your research project allow.

Thank you @ Dmitri Kireev it was helpful and additionally, I agree with your thumb rule.