From owner-chemistry@ccl.net Sat Jul 3 02:45:00 2010 From: "Kalju Kahn kalju**chem.ucsb.edu" To: CCL Subject: CCL: Energy minimization of protein units Message-Id: <-42238-100703023402-4510-hEB99NJTILi2NMW/jy0j9Q^^server.ccl.net> X-Original-From: "Kalju Kahn" Content-Transfer-Encoding: 8bit Content-Type: text/plain;charset=iso-8859-1 Date: Fri, 2 Jul 2010 23:33:53 -0700 MIME-Version: 1.0 Sent to CCL by: "Kalju Kahn" [kalju..chem.ucsb.edu] Hi Jacqueline, I think it is common not to minimize protein structures prior to docking if the protein structure is experimentally available at high resolution. The two main arguments against minimization that I can think of are: 1) It is not clear that molecular mechanics minimization improves upon the experimental structures. The experimental atomic positions are good enough for calculating non-bonded interactions with ligand if the resolution is decent (1.8 angstroms or better); the atomic charges come > from a force field anyways. We may be behind times when computer was an effective denaturing agent but minimizing an good experimental structure is unlikely to make a significant impact on final docking results. 2) The minimization of the ligand-free protein may change structure in an undesirable way ... maybe a sidechain protrudes into the empty binding cavity by a fraction of the angstrom, preventing the binding. In reality, the chain would back away as the ligand binds, but in your docking the protein is most likely treated as a rigid body. There are some situations where (partial) minimization is justified. For example, flexible loops are often not seen in X-ray structures. One could then model-build the missing residues, and perform minimization or restrained MD of this part to minimize steric clashes. I have found NAMD reasonably easy for this. Another scenario is a protein that crystallizes in a tight lattice and there is a reason to believe that crystal contacts deform the binding site (e.g. you cannot get the native ligand to dock). Keep in mind that minimization makes only local changes; in order to hop from one conformation to another you would need MD or MC-type methods. I recall that some commercial software, such as Schrodinger Suite, includes protein minimization as part of the docking work-flow but have not bothered to look if it makes a meaningful difference in results. Hope this helps, Kalju > > Sent to CCL by: "jacqueline cawthray" [jcawth01|,|gmail.com] > We are investigating the rational design of ligands for specific > protein-ligand > interactions. We are quite new to theoretical work on large molecules and > are > not sure how to minimize the protein structures prior to docking > investigations. Is this something Amber or Gromacs is useful for? We were > then > going to use Autodock to investigate various ligand interactions with the > protein. > > Thanks> > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Dr. Kalju Kahn Department of Chemistry and Biochemistry UC Santa Barbara, CA 93106 From owner-chemistry@ccl.net Sat Jul 3 05:25:00 2010 From: "Kalju Kahn kalju]^[chem.ucsb.edu" To: CCL Subject: CCL: Inhibition or Activation Message-Id: <-42239-100703023917-15402-pCM3kiOP07ZRWnqrsCr+yQ _ server.ccl.net> X-Original-From: "Kalju Kahn" Content-Transfer-Encoding: 8bit Content-Type: text/plain;charset=iso-8859-1 Date: Fri, 2 Jul 2010 23:39:09 -0700 MIME-Version: 1.0 Sent to CCL by: "Kalju Kahn" [kalju()chem.ucsb.edu] Carlos: 1) If the protein is enzyme and the ligand blocks the active site (or the cofactor binding site), it is an inhibitor. 2) If the protein is a receptor, there are often known structure-activity relationships; compare your ligand with known agonists and antagonists. 3) Perform the experiment Kalju > > Sent to CCL by: "Carlos Ortega" [caoz96*o*yahoo.es] > Dear CCL users > The reason for this question is about the doubt when realice a molecular > docking in proteins. > > How i know if a ligand docked have power of inhibition or activation in > the protein target?> > > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Dr. Kalju Kahn Department of Chemistry and Biochemistry UC Santa Barbara, CA 93106 From owner-chemistry@ccl.net Sat Jul 3 08:16:01 2010 From: "Vincent Leroux vincent.leroux _ loria.fr" To: CCL Subject: CCL: Energy minimization of protein units Message-Id: <-42240-100703052125-10138-DO3EN61pPIw8MoA1NreuPA|,|server.ccl.net> X-Original-From: Vincent Leroux Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=ISO-8859-1 Date: Sat, 03 Jul 2010 11:21:12 +0200 MIME-Version: 1.0 Sent to CCL by: Vincent Leroux [vincent.leroux++loria.fr] Hi Jacqueline, Most molecular modelers will say you should never mess with good quality structural data unless you know exactly what you are doing. However, with large biomolecules we live in a world of models, more or less closely derived from experimental data. Models should always be challenged. If you have a X-ray structure for your protein, you should have in mind that the data was *already* optimized during the X-ray structure determination refinement process, using specialized approaches for doing so. You are not going to improve this using a simple minimizer - quite the contrary indeed. I would suggest you doing so anyway, especially if you are new to the field. Minimize and then equilibrate and do a MD simulation for good measure, in implicit (MM/PBSA) or explicit solvent. Provided the program does not crash miserably (as it will probably if not properly used), if the whole structure gets screwed up quickly (e.g. secondary structure elements get destroyed), that means (1) you did wrong with the simulation setup (and the sooner you realize where you made the mistake the better) or (2) the model is not so good... and yes, CHARMM, AMBER, GROMOS, NAMD and the like are what you need then. If you do some serious structure-based drug design work you will need them at some point anyway. Important: when strange things happen, if you started from a theoretical model, check (1), then (2). If you started from a good X-ray structure, check (1), then (1) again... and again... and if you are *sure* you made no mistake (e.g. your parameters do work perfectly with some other random hi-res X-ray structures) check G.Kleywegt's papers on X-ray structures validation. Maybe you got a lousy one, this happens. If you succeed, then you can use the starting structural data for docking. It is most probably reliable, and you learned how to so simple molecular mechanics simulations in the process :-) Do not use the data you generated, unless you are going to continue performing simulations on the system using the same force field, or the simulations did reveal some interesting things (in such a case you will probably use some derived target conformations for docking in addition to the reference one). In any case, you will have to model things on your target protein. From the easiest task to the hardest: protons, histidines protons, protonation state of proteins not operating at pH ~7, single-point mutations, solvent, partly or totally unresolved residues, missing segments (probably loops), induced fit effects. Golden rule: what you model, you have to minimize. If you did well, minimization will converge very quickly. If you modeled large things, in addition you should check the stability of your system. MD in implicit (MM/PBSA) or better explicit solvent is the method of choice. Good luck! Regards, VL Le 03/07/10 01:06, jacqueline cawthray jcawth01*|*gmail.com a écrit : > Sent to CCL by: "jacqueline cawthray" [jcawth01|,|gmail.com] > We are investigating the rational design of ligands for specific protein-ligand > interactions. We are quite new to theoretical work on large molecules and are > not sure how to minimize the protein structures prior to docking > investigations. Is this something Amber or Gromacs is useful for? We were then > going to use Autodock to investigate various ligand interactions with the > protein. > > Thanks> > >