From tudor@wucmd.wustl.edu Sun Feb 6 23:50:17 1994 Received: from wugate.wustl.edu for tudor@wucmd.wustl.edu by www.ccl.net (8.6.4/930601.1506) id WAA15056; Sun, 6 Feb 1994 22:59:08 -0500 Received: by wugate.wustl.edu (5.67b+/WUSTL-0.3) with SMTP id AA04705; Sun, 6 Feb 1994 21:59:08 -0600 Received: by wucmd (911016.SGI/911001.SGI) for @wugate.wustl.edu:CHEMISTRY@oscsunb.ccl.net id AA03240; Mon, 7 Feb 94 02:03:34 GMT Date: Sun, 6 Feb 1994 18:35:52 -0600 (CST) From: Tudor Oprea Subject: Re: CCL:rational drug design of HIV-protease inhibitors To: Marty Gallagher Cc: CHEMISTRY@oscsunb.ccl.net In-Reply-To: <9402041810.AA14007@ionchannel.med.harvard.edu> Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII On Fri, 4 Feb 1994, Marty Gallagher wrote: > Hello, > > A while back someone asked about success stories in rational > drug design. In that light, what does the group think of the Merck > groups article in *Science* 23:380-384 where they rationally designed > inhibitors of HIV protease? > > -- Marty > > > ======================================================================== > Martin J. Gallagher phone: (617) 432-1729 > Dept. of Neurobiology fax: (617) 734-7557 > Harvard Medical School E-mail: marty@ionchannel.med.harvard.edu > 220 Longwood Ave > Boston, MA 02115 Background information: here at Washington University we developed a 3D-QSAR - CoMFA model for 59 HIV-protease inhibitors (see J Med Chem 36:4152-4160 1993) - a model that was based on experimentally derived alignment rules (crystals of 7 inhibitors) and was tested for predictive power (36 compounds with different chemistry - but peptides as the other 59) and explanatory power (CoMFA fields were compared and interpreted in the light of binding site residues in immediate contact with the inhibitors). The work on predictive power was submitted to J Med Chem, and the one on explanatory power is to be printed in Drug Design and Discovery. Statistical results: cross-validated r^2 = 0.778 (6 PCs) - 59 cmpds, leave-one out procedure "press" = 0.552 r^2 = 0.984 (6) s = 0.146 F-test = 549.838 Contributions: .36 steric .64 electrostatic predictive r^2 = 0.679 (34 cmpds - 2 outliers excluded after factorial analysis) all compounds neutral The non-peptide cyclic ureas were announced at the Gordon Conference in QSAR (August 1993) where Dr Eyermann gave a lecture (no activities, though). The four compounds with published activities were predicted by our model as follows (see the Science paper for #s): Cmpd Actual Predicted (both as IC50, micromolar) 1 0.63 0.132 2 0.30 0.140 3 0.22 0.23 4 0.036 0.061 (the only error higher than .3 microM or .47 kcal/mol in binding affinity is for cmpd 1 - the above error is the one of our model tested on 34 cmpds; also note that x-tal "data" is available only for cmpd 3, as Fig 3 in the Science paper). Comments regarding the drug design strategy used by the DuPont-Merck group: 1. it is based on a simple concept: design rigid compounds (or as rigid as you can), because you will not have to pay for entropy, hence you gain potency 2. the cyclic urea framework that forms the basis of these inhibitors was designed to maximize interactions with the catalytic aspartates and other key residues in the binding site, and also to replace a catalytic water - all this while providing an elegant solution to the problem of transition state isostere and in the same time 3. the fact that these are non-peptide compounds makes them orally available (an essential target for any pharmaceutical company) 4. using specific structural information for design makes these compounds specific for HIV protease - they do not inhibit (except at very high concen- trations) cellular aspartic proteases Comments regarding the drug design tactics used by the DuPont-Merck group (or how did they 'set forth' to get advantages in the HIV-protease front): structural knowledge of the free enzyme and the inhibitor-bound enzyme was ESSENTIAL for this study, because specific targets were aimed at: * atoms in the binding site which were important in the economy of binding which were used to define the * 3D pharmacophore (exact location of binding site atoms - which are then used as template for putative structures that can match this pattern), which was then used as a query in a * 3D database search (useful concept: has anyone done such a compound before? - ... or how close to our 3D pattern are already-made-cmpds ?) this proves many times to be of help, but users are aware that such a method would not suggest NEW compounds, and this is where the molecular modeler (or was it a medicinal chemist?) uses knowledge blended with imagination to come with the * "_initial_idea_ for a nonpetide inhibitor that includes a structural water mimic" (exact quote) - this suggests that someone in the group came with this suggestion, upon which the work was continued later on, the skeleton was refined to ensure * better complementarity with the 3D pharmacophore (e.g., two hydroxyls instead of one, to bind two aspartates in the catalytic site, and the urea instead of the keto to improve electrostatic negative pattern in the flaps region for better hydrogen bonding) Predictions using our CoMFA model suggest (in a certain way) that the published compounds are consistent with the peptides used to generate our CoMFA model: the right steric and electrostatic features were placed at similar location. The insertion of the CoMFA results in this comment - which may seem inappropriate - is merely to suggest that another technique, 3D-QSAR, is available in the process of drug design (although it was not used for the work published by the DuPont-Merck group). The Science paper proves that current concepts used in drug design are valid - but it is quite clear that a scientist (or a team of scientists) is needed, because this problem was not solved entirely by a computer. However, the use of computational methods was essential. ========================== Tudor-Ionel Oprea = Tel. (1-314) 935 4672 = Research Associate = Fax. (1-314) 935 4979 = Washington University Center for Molecular Design = Email: = Lopata 510, Box 1099, One Brookings Drive = tudor@wucmd.wustl.edu = St.Louis MO 63130 ========================== Note: no philosophical statements, cartoons, humor or disclaimers below