Have a look at the full crystal repeat of the PDB structure if you can . If a protein molecule in an adjacent unit cell is making contacts around the active site in a non-biologically relevant way it can sometimes prevent a ligand from sitting in the normal orientation. This crystallographically generated artifact is known in a number of cases and is certainly under-recognised in others.

If ligand stays there and have significant differences in pose and RMSD from original known binding mode then yes you are right. Most possibly you got the alternative binding mode of your ligand. You can validate and check the same by simple flexible docking experiment by using final MD ligand coordinates as reference with final docked pose suggested by docking program.

Two different methods suggesting the same reflects the correctness of your approach and result.

What do you mean with "ligand stays happily inside the active site pocket" ???

Hi Dario,

During the long simulation run (50ns) the ligand is staying in the active site without much change in the binding pocket and the rmsd of the protein.

Hi, I would go by docking in order to sample as much as you can the binding site. then you can score the poses and compare the results with the known bound structures. the best scored pose then you can perform MD.

Doing MD straightforward is difficult to see different binding mode if you are in a local minimum. 50ns is a very short time for this to happen. have look to "rare event" in literature.

hope it helps

cheers

Hi Tarak,

I think Andrea above has a good point. Run a series of docking studies and sample the poses via MD to compare the averaged RMSD and binding energies. That should help you to explore the potential binding mode space more thoroughly.

Additionally, you might want to consider including an analysis of the ligand-binding interactions of the poses to identify those with either unfavourable, or unsatisfied interactions. There's some good review studies out there that suggest this analysis in conjunction with docking scores is a better strategy and can help to differentiate true actives from decoys, etc. Here's a couple of useful references.

DOI: 10.1021/ci300200r

DOI: 10.1021/ci300566n

I'd recommend you do the non-bond analysis before the MD simulations. The results can indicate where you may either have to re-sample protein side-chain conformations, may need to consider explicit waters, or, may give a strong hint to the pose being undesirable.

If you don't have access to an extensive set of favourable, unfavourable and unsatisfied non-bond monitors, you can download the free DS Visualizer from Accelrys. Note: I work for Accelrys.

http://accelrys.com/products/discovery-studio/visualization-download.php

Best regards,

Adrian.

You could check

1. Binding energy differences of the two observed modes

2. Flexible docking of protein environment and ligand

3. Modifications in critical residues in P-L-I differences through mutation studies

4. Any conserved waters- ligand interactions...

If all of the results have general consensus, you may propose the alternate conformation that can be verified by structural or biochemical studies.. goo d luck :-)

Thanks to all for your valuable and nicely explained suggestions. I'll try to proceed accordingly.

Dear Tarak.

Although two different binding modes are theoretically possible, it is really difficult to find two significant binding poses for the same ligand in the same site with similar affinity. This is due to an exponential dependence of affinity with binding energy.

During a MD simulations a local minima can be found and vibration/oscilation around a binding site do not necesarily represent a good docked ligand. Suggestions above studying your ligand docking, or analysis of the binding site interactions with your poses obtained from MD are good strategies. But a questions arises, hoy different are your binding orientations, because maybe they are found by relatively small rearrangements of the ligand and of the side chains from the binding sites. In this case I consider more probable the existence of different orientations with similar affinities, or interconversion of those binding modes.

Good luck !

Arturo.

Dear Sir,

Thanks for the reply. The new binding mode has been observed in one of the crystal structure of a protein from a similar species.

The orientations of the ligand in highly different. It's about 60 deg rotation along the molecular plane of the ligand in the active site.

Affinities of the ligand in two different orientations are not known upto now.

Have a look at the full crystal repeat of the PDB structure if you can . If a protein molecule in an adjacent unit cell is making contacts around the active site in a non-biologically relevant way it can sometimes prevent a ligand from sitting in the normal orientation. This crystallographically generated artifact is known in a number of cases and is certainly under-recognised in others.

I agreed with Andrea and Adrian. To me, the easiest way to move forward is to use dockings to explore protein-ligand complexes. You may have different orientations with small difference in energy or have some geometric restriction imposed by the crystal (something in the line of what John says). Protein-ligand docking should help you a lot.

Since you already have performed MD runs, you may want to perform the dockings on an ensemble of conformations of the receptor (i.e. snapshot or cluster representatives).

Good luck,

JD

As discussed above, MD is not a great way to predict the binding mode of a ligand in your pocket. There are enough good arguments that have been discussed above.

PDB data in a similar protein can be a very good start in determining your binding mode, but it will depend on the similarity/identity between your protein and the determined structure. All usual considerations regarding structure quality and protein conformation (is it in a relevant conformation, i.e. active conformation?) apply. John's comment regarding packing artifacts is also very relevant!

Next to theoretical considerations regarding docking & MD, I'd spend some time on the SAR within the compound series, if you have some data available, try to see if your binding mode(s) make sense with respect to what is tolerated (or not) within your series. This exercise is usually valuable!

For me if you want to see different binding mode of the ligand bound to a particular protein relatively in a short simulation time in that case you have to do reconnaissance metadynamics (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3325676/). 

By this method you can explore different binding poses of the ligand inside certain protein structure. Suppose if your ligand has pose, x1 inside the protein y and it is reported experimentally. But it is possible that ligand is more stable in a different binding pose which is not explored by crystallography. In that case you can explore those binding poses such as x2, x3, x4 etc. using reconnaissance metadynamics. Suppose if pose x2 remains in a particular metadynamics basin for a longer period of time and displayed a favorable free energy surface as compared to crystallography and other binding poses in that case you can definitely say that pose x2 is more stable than the crystallographic one.

Suppose you have two crystallographic pose of same ligand say for example pose x11 and x2 and both crystallographically identified and from FEP simulation you identified that pose x2 is more energetically favored when compared to x1. In that case using reconnaissance metadynamics you can possibly mimic to the binding pose of x2 in a very short timescale with great accuracy.

Hope it helps. But we are trying to develop it further for these type of applications.

Hope it helps,

SB