From arthur@csb0.IPC.PKU.EDU.CN Sun Apr 21 08:48:00 1996 Received: from csb0.IPC.PKU.EDU.CN for arthur@csb0.IPC.PKU.EDU.CN by www.ccl.net (8.7.1/950822.1) id IAA03772; Sun, 21 Apr 1996 08:32:19 -0400 (EDT) Received: by csb0.IPC.PKU.EDU.CN (920330.SGI/940406.SGI.AUTO) for chemistry@www.ccl.net id AA01920; Sun, 21 Apr 96 20:29:00 -0700 Date: Sun, 21 Apr 1996 20:29:00 -0700 (PDT) From: Arthur Wang To: CCL mailing list Subject: Summary: Quantitative prediction of binding affinities Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear CCLers, Last week I posted the question about prediction of binding affinities. Several answers have come into my mailbox. They are proved to be very helpful. I thank all of these concerns and the summary is here. My original question is: --------------------------------------------------------------------------- In structure-based drug design, maybe the real problem lies in scoring the candidates, i.e. predicting the binding affinities to find the promising ones among the potential ligands. Obviously if the score relates directly to the free energy changes upon the binding process, designers will be lent much convenience. Many approaches have focused on this problem. Some use rule-based energy functions (e.g. J.Comp.-Aided Molecular Design, 1994, 8, 243-256), some use force field to evaluate the energies (e.g. GRID, DOCK, & J.Comp.Chem., 1995, 16, 454-464). For some reasons, we prefer force field method. (Any comments on the comparison of these two methods are welcome! :-) In such a energy function, these terms are usually included: E = E(vdW) + E(electrostatic) + E(desolvation) + E(H-bond) + ... Different force fields often have different forms. So here comes my question: Are there anyone who has checked the validities of these force fields? Which one is more reasonable and which is less reasonable? Usually a few crystal structures of complex are rebuilt by using a specific method, but so far I have not seen any attempts to predict quantitatively the experimentally determined binding constants or free energy changes. This is critical. Because we will probably yield nothing valuable by using a false energy estimation. Any comments? Any clues? I would like to calculate the binding energies for a series of complexes by using current force fields. In this case, I need a list of complex PDB entries which have reliable determined binding constants. Anyone maintains such a list? ------------------------------------------------------------------------------ ANSWERS: >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> From: "Dr. Chris L. Waller" A group of us in Garland Marshall's lab at Washington University were very interested in this particular problem. To this end we developed a "force field" which was designed to predict the Ki's for untested compounds. To my knowledge, this manuscript is in press in JACS. Look for a manuscript entitled "VALIDATE" by Head, et al. >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> From: hou@agouron.com (Xinjun Hou) >Are there anyone who has checked the validities of these force >fields? We asked similar question a few years back. One motivation was to find a relative efficient method to predict binding energy, which can be used in our daily drug design project. Based on our own ligand-protein cocrystal structure databases, we found that a generic force field could give reasonable relative binding energy for cogeneric series of ligands. Some sort of solvation model is needed to obtain better correlation. >...but so far I have not seen any attempts to predict quantitatively >the experimentally determined binding constants or free energy changes. No really correct, a good review can be found in Ajay and M. A. Murcko, J. Med. Chem. 1995, V38:N26:P4953 >...I need a list of complex PDB entries which have reliable determined >binding constants. Check out a publication from G. R. Marshall's group (could not find the reference, might be ref#80 in Ajay's paper) >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> From: "W. Todd Wipke" There are many examples showing that one can not predict the binding affinity with molecular mechanics by only considering the substrate in the protein cavity. It is only part of the story. -Todd Wipke >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> From: CHEM193@csc.canterbury.ac.nz Dear Arthur; I read your post in the CCL, it is realy frastrating as I met it few years ago. There are many force fields for the energy calculation in drug design, predicting protein and DNA, no people know which one is best or worst, the choice of the force field is subject to your own purpose. Try the force fields in your calculation to see which one is the best to fit your problem. In protein/DNA prediction, there are three force fields are commonly used, ECEPP/2 which was developed by Sheraga etc, Amber and Opls, also a combination of the AMBER/OPLS in Macromodel package is good for the prediction of the proteins. >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> From: PEARLMAN@VAX.PHR.UTEXAS.EDU Dear Arthur -- > energy function, these terms are usually included: > > E = E(vdW) + E(electrostatic) + E(desolvation) + E(H-bond) + ... > > Different force fields often have different forms. So here comes my > question: Are there anyone who has checked the validities of these force > fields? Which one is more reasonable and which is less reasonable? Usually > a few crystal structures of complex are rebuilt by using a specific > method, but so far I have not seen any attempts to predict quantitatively > the experimentally determined binding constants or free energy changes. > This is critical. Because we will probably yield nothing valuable by > using a false energy estimation. There have been a few studies comparing molecular mechanics force fields in terms of how well they reproduce experimentally determined geometries and rotational energy barriers but I am not aware of any studies attempting to validate standard molecular mechanics force fields for the purpose of calculating *inter*molecular interaction energies. Since the force fields were not developed for this purpose and were not parameterized for this purpose, they will not be particularly accurate when used for this purpose. I'm glad to see that you included E(desolvation) in your model. Actually, as you probably know, it should be Delta-G(desolvation). Have you also thought about the need to address the loss of conformational entropy which occurs as ligands with rotatable bonds become "locked" within the receptor? Also, don't forget to consider the change in conformational energy which occurs as the ligand distorts from its preferred conformation in solution to the conformation imposed by the receptor. Since you will have a difficult time guessing the exact conformation imposed by the receptor, there will be considerable error associated with this contribution! Perhaps so much that other errors don't seem so important. Perhaps so much that the notion of a scoring function becomes less "exciting." -- Bob Pearlman >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> From: "Alexander Hillisch" Have a look at: J.Med. Chem. (1995) 38, 305-317 People form Merck describe molecular mechanics calculations with they modified MM2 force field to predict activities of HIV-PR-inhibitors. I would be interested in other responses to your question to CCL. I am doing similar things so I would like to please you for closer contact with me. >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Regards, Arthur _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ _/ Arthur Wang Doctoral Candidate _/ _/ Molecular Design Lab _/ _/ Institute of Physical Chemistry, Peking University _/ _/ Beijing 100871, P.R.China _/ _/ _/ _/ E-mail: arthur@ipc.pku.edu.cn _/ _/ Tel: 86-10-2751490 Fax: 86-10-2751725 _/ _/ WWW: http://www.ipc.pku.edu.cn/moldes/arthur/home.html _/ _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/