Dear Anuj,

The majority of the above comments are correct but from active users of Molecular computational methods. My background is both crystallography and computation. So I have more critical view of the processes used.

There are two aspects that are not addressed above. Every PDB deposited structure I have looked at so far contains conformational and solvent errors. There are no proper tools that would address this problem. You can go to validation tools mentioned above but this is not a solution.

On the other hand a vast majority of molecular software tools are too crude (in particular all the MD packages based on classical mechanics assumptions) to provide real atomic details of your molecular system. In my hands proteins that were refined at higher than 1Å resolution diverged in different directions from the natively refined structure when subjected to minimization alone and to MD alone.

The advantage of molecular simulation tools are that you may spot and partially fix conformational problems associated with the X-ray or NMR structures. They will inevitably lead you away from the experimental structure but will restore the most likely conformations. The other advantage of simulation tools is that particular X-ray structures do not capture well the mobility of the proteins. However, which particular tool to use is really a matter of taste not a demonstrated superiority so it does not matter. Use of docking tools matter much more.

Additional element to consider is that in your particular application (docking and screening) the atomistic accuracy do not count that much. Therefore if you try to follow the advice given by Girdhar (fix what you can fix, including adding missing elements, and then MD and minimization) you may get to reasonable answers. At this stage make sure that whatever crystallographers claim to have is true as the newest studies showed that small molecule ligands are the most frequently erroneous in the structures.

Dear Anuj, In my opinion you should go for preliminary treatment like adding hydrogen, adding missing residues, refining the loop and finally energy minimization. It will allow CADD software to recognize/treat the PDB protein coordinates in a better way. Some times, without minimization the PDB structural coordinates creates problems in further steps/process.

Thanx sir, would you please suggest me some freewares ?

You should follow as per Mr Giridhar. He is right that we need to preprocess the protein for minimizing any erros or strains.

For energy minimzation it is well practised method to run a steepest descent minimization algorithm followed by conjugated gradient. Free softwares to do these is Gromacs.

You can check any model validation programs such as Procheck (use PDBsum), Errat and verify3d, prosa to check whether minimzation resulted into a bad structure? These program gives u both local and global quality of model.

You can always measure RMSD of minimzed structure compared to your original model to see whether RMSD resulted in a drastic change of protein backbone or not. For RMSD u can use any structural alignment programs (Swiss PDB viewer, DALI, VAST)

It depends on how you choose to use the structure. If you are trying to reproduce the binding mode, don't minimize the whole thing. Just go through the steps to prepare the structure and optimize the water positions and hydrogens. Of course, you should run a protein health utility to determine if the structural biologist's work is good, paying particular attention to close contacts and Ramachandran outliers. Those definitely need to be corrected, at least locally. But if they are distant from the binding site, I wouldn't sweat them too much.

I have found that minimizing the structure sometimes changes the binding mode. if you want to see if this is true for your structure, minimize it with oiginal ligand and do a RMS afterwards. You want to make sure that all of your non-bonded interactions are preserved before you proceed. If you minimize and positions in the binding site change significantly, you will need to evaluate the binding mode. I generally use Schrodinger's Maestro with OPLS. It's an acceptable force field for both proteins and ligands.

Definitely yes, for the reasons given in the previous answers - and you will probably want to build hydrogen positions (as mentioned by previous responders).

In my opinion loops should not be optimized but you should add hydrogens (protonate the structure). The structure should only be minimized when your application involve Molecular Dynamics Simulations, otherwise the structure in itself sufficient for primary applications.

I agree with the comments above. An additional concern is the source of data. If 3D structure is from x-ray, definitely YES-minimization and also MD simulation at physiological conditions is necessary. If the structure is obtained from NMR, then it may be used directly.

Even for NMR resolved structures in the PDB, if you think about using any force-field based analysis, you go for an initial energy minimization in the conditions of your system. GROMACS (www.gromacs.org) is really one good option as it is really fast, flexible and free as their slogan is. If you're looking into minimizing your ligand for which GROMACS doen't always have topology information built-in, try using ACPYPE which is a python interface working with Ambertools. All are free to download and use and really good softwares.

For standard applications you don't need to minimize. For MD simulations yes you have to minimize after doing the usual addition of Hs, missing residues etc.

Thanks a lot. I am sure all suggestions are valuable. Thanks gentle men.

Many authors perform the minimization others don't so the right way is the working way. You should verify how your docking program works with that receptor's structure: try to validate the procedure by re-docking inside the binding site the crystallographic ligand (if you have one) in the original and in the minimized receptor's structure. The way that reproduce better the crystallographic complex (lowest rmsd between crystallographic and predicted ligand's position) is the way you should use.

Hi anuj. If your PDB structure is X-ray crystal structure, addition of H's and then minimization should be done. You can use Molecular Operating Environment (MOE) for minimization process.

Its batter to minimize structure using Amber3 or GROMACS

i think that is better to minimize structure use OPLS_2005 of schrodinger, because MOE is more abrupt with minimization, the same way happen with amber and charmm.

Schrodinger -OPLS is available with me, I tried using chimera ( steepest decent and conjugant gradient) Some one pls suggest me preferable parameters if I use chimera, I am trying for gromacs too. :)

Always minimize. I will minimize the raw structure first, and with numerous cycles if I have removed ligands or added in missing residues or heavy atoms etc. I also minimize the structure after it has been solvated or whatever medium has been added. I will minimize the environment around the protein structure, then the protein on it's own, and then the whole system in turn. I prefer to do a mixture of steepest descent and conjugate gradient. The steepest descent algorithm can get trapped in local minima. The adapted basis newton raphson (ABNR) algorithm is superior though; it's present in CHARMM, I'm not sure about the other packages. If you want more details about how the algorithms compare and the drawbacks of each one etc. to assist you in making an appropriate choice then I would recommend you have a look at Leach's textbook. This book helped me immensely in understanding and selecting the protocols most suitable for what I was studying.

http://www.amazon.com/Molecular-Modelling-Principles-Applications-2nd/dp/0582382106

It depends on the studies you are performing

Dear Anuj,

The majority of the above comments are correct but from active users of Molecular computational methods. My background is both crystallography and computation. So I have more critical view of the processes used.

There are two aspects that are not addressed above. Every PDB deposited structure I have looked at so far contains conformational and solvent errors. There are no proper tools that would address this problem. You can go to validation tools mentioned above but this is not a solution.

On the other hand a vast majority of molecular software tools are too crude (in particular all the MD packages based on classical mechanics assumptions) to provide real atomic details of your molecular system. In my hands proteins that were refined at higher than 1Å resolution diverged in different directions from the natively refined structure when subjected to minimization alone and to MD alone.

The advantage of molecular simulation tools are that you may spot and partially fix conformational problems associated with the X-ray or NMR structures. They will inevitably lead you away from the experimental structure but will restore the most likely conformations. The other advantage of simulation tools is that particular X-ray structures do not capture well the mobility of the proteins. However, which particular tool to use is really a matter of taste not a demonstrated superiority so it does not matter. Use of docking tools matter much more.

Additional element to consider is that in your particular application (docking and screening) the atomistic accuracy do not count that much. Therefore if you try to follow the advice given by Girdhar (fix what you can fix, including adding missing elements, and then MD and minimization) you may get to reasonable answers. At this stage make sure that whatever crystallographers claim to have is true as the newest studies showed that small molecule ligands are the most frequently erroneous in the structures.

dear all and respected Boguslaw, i want to add on few things and clarify few things. Mostly i agree with sahil's and Kimberly's approach and the reasoning is as follows

as far as i know the high success of re-docking, the primary test for a good docking protocol, is based on the fact that pocket is adapted to ligand present in the crystal structure. Leave minimisation even the same ligand downloaded from a database such as pubchem can give different pose on docking. Hence, minimising the complete protein which as stated by Boguslaw will lead away from the crystal structure will disrupt the pocket and surely will give poor re-docking and cross docking can be even worse. what i have been doing till now is addition of hydrogens and minimising them only. As docking is mostly rigid in nature hence missing residues outside active site should not be of any concern.

For MD simulation, which is more dynamic approach, i do add missing atoms, minimize structure before production run, This makes sure that any interaction lost due to missing residues can influence the interactions close by and ultimately propagating through whole protein unless the residues are at very end.

Now, as far validation tools to check goes i am not sure which one to use.can you tell me which is best possible option to validate structure as Boguslaw said there are no proper tools.

I agree with the comments of Boguslaw and Nitin. The most important thing is to check your protein for obvious errors, especially around the active site and correct them, manually if necessary. Always check against the crystal structure electron density to be sure that residues and ligands are truly where they are supposed to be. Always check that histidines, asparagines and glutamine side chains are correct orientation or tautomer - the electron density can't tell you this (some software such as MOE can do these checks automatically but you should still also check by eye ). Minimisation will not correct any of these type of errors .

If your model is free of these type of errors it may be good enough to dock to without any further work. However If, after making these corrections, I were to minimize the protein i would do so with restraints on all heavy atom positions. Otherwise the model you end up with may be quite different in important respects to the X-Ray structure. Then you are docking into a theoretical structure not one that corresponds to the experimental evidence.

for the tautomers and correct ionization you can use pdb2pqr webserver

http://nbcr-222.ucsd.edu/pdb2pqr_1.8/

and assign the protonation states according to the concerned pH

for Missing residues you can check REMARK 465 of the PDB file and REMARK 470 for Missing Atom. if these remarks are not present in your PDB file then you don't need to worry a lot.