From owner-chemistry@ccl.net Fri Aug 4 04:01:14 1995 Received: from danpost.uni-c.dk for peon@medchem.dfh.dk by www.ccl.net (8.6.10/930601.1506) id DAA05895; Fri, 4 Aug 1995 03:52:46 -0400 Received: from medchem.dfh.dk (medchem.dfh.dk [130.225.177.15]) by danpost.uni-c.dk (8.6.4/8.6) with ESMTP id JAA07659; Fri, 4 Aug 1995 09:52:42 +0200 Received: from [130.225.177.59] by medchem.dfh.dk via SMTP (940816.SGI.8.6.9/940406.SGI) id KAA28951; Fri, 4 Aug 1995 10:24:30 +0200 Message-Id: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Fri, 4 Aug 1995 09:52:43 +0200 To: BILL WELSH From: peon@medchem.dfh.dk (Per-Ola Norrby) Subject: Re: CCL:Question about Electrostatics in Molecular Mechanics Cc: chemistry@ccl.net At 16.34 95-08-03, BILL WELSH wrote: >Dear Netters, > >Please help settle a small controversy. For a particular molecular >mechanics calculation, I derived partial atomic charges from bond >dipole moment measurements made in nonpolar solvent. I then ran >molecular mechanics calculations using an in vacuo dielectric constant >of 1.00. A colleague argues that I must use a dielectric constant >appropriate to the nonpolar solvent from which the charges were >derived (e.g., 2-3). It seems to me that the charges already embody >the effect of the nonpolar solvent's dielectric constant, and so using >the in vacuo value of 1.00 is okay and perhaps preferred. Stated another >way, it seems to me that using these charges AND the dielectric of 2-3 >would count the effect of the nonpolar solvent twice. > Dear Bill How to handle charges correctly is probably one of the hardest problems in molecular mechanics today. The recommendations usually vary between the force fields too, so some details on your system would help. However, almost all mature methods use environmentally independent partial charges (or bond dipoles, ...). The parameters of your force field are probably derived to reproduce gas phase data, so you should expect best results if you also want to predict gas phase or in vacuo behavior. If you want to predict how the molecules behave in a solvent, you should add a solvent model. The simplest possible solvent model is changing the dielectric constant. So, what are you trying to predict with your calculations? What force field are you using? Best Regards, PeO Norrby * Per-Ola Norrby * The Royal Danish School of Pharmacy, Dept. of Med. Chem. * Universitetsparken 2, DK 2100 Copenhagen, Denmark * tel. +45-35376777-506, +45-35370850 fax +45-35372209 * Internet: peon@medchem.dfh.dk, peo@compchem.dfh.dk From owner-chemistry@ccl.net Fri Aug 4 06:31:16 1995 Received: from tigris.klte.hu for tamasgunda@tigris.klte.hu by www.ccl.net (8.6.10/930601.1506) id GAA06772; Fri, 4 Aug 1995 06:17:39 -0400 Message-Id: <199508041017.GAA06772@www.ccl.net> Received: from anti02 (anti02.chem.klte.hu) by tigris.klte.hu (MX V4.1 VAX) with SMTP; Fri, 04 Aug 1995 11:12:14 EDT Sender: X-MX-Warning: Warning -- Invalid "From" header. From: "tamasgunda@tigris.klte.hu" To: chemistry@ccl.net Date: Fri, 4 Aug 1995 11:12:15 +1 MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7BIT Subject: principal components & least-square proc. Priority: normal X-mailer: Pegasus Mail/Windows (v1.22) Dear Netters, I have got several answers to my yesterday question on principal components. I will send a summary for those who are interested in the problem. However there is a point I should shed more light to. The problem in question is, as several people realized, the construction of a best-fit plain to points in the 3D space. Some answers suggested that this is an ordinary least-square fitting problem in 3D. It is not so simple. It would be very easy to explain it with pen & paper, I try now in words. Imagine a normal least-square procedure: we have a number of experimental data, and we construct the linear regression equation y=ax+b. So we have the points in the x-y system, and the regression line: y | ./ | . / | . / | / | / | ./ . | / . | ./ |_________________________________________x (If the illustration looks awfull, switch to the fixed system font or fixed courier) According to the math of the normal least square procedure, as far as I know, NOT the distances of the points and the line are minimized (i.e. a perdendicular from a point to the line), but the difference of the measured and the calculated y values, which is represented graphically by a line between the point and the regression line and it is parallel with the y axis: y | / | . / | | / | | / | |/ | / | / | |_________________________________________x (Sorry, if my English is similar to the flight of a penquin, but I hope it can be understand.) The same applies in 3D,4D etc situations. One of the consequences is for example that if I rotate the data set, the regression line (plane) will not rotate accordingly, but funny results can be obtained. This is the reason why I asked for help concerning the principal component method in my yesterday e-mail. (Although I like math, I am a chemist first and foremost) Best wishes Tamas ***************************************************************************** Tamas E. Gunda, Ph.D. phone: (+36-52) 316666 ext 2479 Research Group of Antibiotics fax : (+36-52) 310936 L. Kossuth University e-mail: tamasgunda@tigris.klte.hu POBox 36 H-4010 Debrecen Hungary ***************************************************************************** From owner-chemistry@ccl.net Fri Aug 4 07:16:17 1995 Received: from btm2d1.mat.uni-bayreuth.de for wieland@btm2d1.mat.uni-bayreuth.de by www.ccl.net (8.6.10/930601.1506) id HAA07457; Fri, 4 Aug 1995 07:15:10 -0400 From: Received: by btm2d1.mat.uni-bayreuth.de (IBM OS/2 SENDMAIL VERSION 1.3.2)/1.0) id AA0130; Fri, 04 Aug 95 13:13:59 +0200 Message-Id: <9508041113.AA0130@btm2d1.mat.uni-bayreuth.de> Date: Fri, 4 Aug 95 12:46:49 MET To: chemistry@www.ccl.net, CHMINF-L@IUBVM.INDIANA.EDU, ICS-L@umdd.umd.edu, orgchem@extreme.chem.rpi.edu, physics@ubvm.cc.buffalo.edu Subject: Graph-theoretical description of mass spectra Dear Netters, we are planning to develop methods for the automatic interpretation of mass spectra. In this context, we are also thinking of describing the spectra by graphs, for example peaks are vertices and the edges reflect some kind of "B comprises A". We would like to get in contact with scientist who have also some experience in this field. Moreover, we would appreciate any pointers to literature on automated interpretation of mass spectra in general and on the usage of graphs for that in particular. I will summarize the answers if there is enough interest. Thanks in advance. Best wishes, Th. Wieland P.S.: The summary of my recent question to CCL on topological indices will come soon. I'm still waiting for some sources to answer. Thomas Wieland +---------------+ Dipl. Math. |+---- +----+| Lehrstuhl II f. Mathematik |\ \ | || Universitaet Bayreuth | \ \ | || | \ \ | || 95440 Bayreuth | \ \\ || Germany | \ \\ || Tel. +49 (921) 553386 | \ \\ || Fax +49 (921) 553385 | \-------|| +---------------+ From owner-chemistry@ccl.net Fri Aug 4 10:01:17 1995 Received: from faui45.informatik.uni-erlangen.de for goeller@organik.uni-erlangen.de by www.ccl.net (8.6.10/930601.1506) id JAA10812; Fri, 4 Aug 1995 09:59:04 -0400 Received: from derioc1.organik.uni-erlangen.de by uni-erlangen.de with SMTP; id AA26275 (5.65c-7/7.3w-FAU); Fri, 4 Aug 1995 15:58:22 +0200 Received: by organik.uni-erlangen.de; id AA22694 (5.64/7.4b-FAU); Fri, 4 Aug 95 15:58:20 +0200 From: Andreas Goeller Message-Id: <9508041358.AA22694@derioc1.organik.uni-erlangen.de> Subject: electronegativity To: CHEMISTRY@ccl.net Date: Fri, 4 Aug 95 15:58:19 METDST X-Mailer: ELM [version 2.3 PL11] I'm searching for values of electronegativities for some functional groups. I'm doing investigations concerning EN vs. bond lenth etc. The ligands in the systems are F OH Cl NH2 Br I CH3 H This would be the "right" order, if you take into account only the EN of the atoms. Some other suggestions or literature? Bye, AHG... --------------------------------------------------------------- Andreas Goeller Computer Chemie Centrum der Dipl.-Chem. Universitaet Erlangen/Nuernberg Naegelsbachstr. 25 phone: +49(0)9131-856583 D-91052 Erlangen fax: +49(0)9131-856565 Germany email: goeller@organik.uni-erlangen.de http://www.organik.uni-erlangen.de --------------------------------------------------------------- From owner-chemistry@ccl.net Fri Aug 4 10:16:17 1995 Received: from bioc1.biosym.com for tony@bioc1.biosym.com by www.ccl.net (8.6.10/930601.1506) id KAA11030; Fri, 4 Aug 1995 10:13:11 -0400 Received: from biosym.com by bioc1.biosym.com (5.64/0.0) id AA26956; Fri, 4 Aug 95 06:10:51 -0700 Received: by east1.biosym.com (920330.SGI/920502.SGI) for bioc1!ccl.net!chemistry-request id AA07441; Fri, 4 Aug 95 08:57:38 -0400 Date: Fri, 4 Aug 95 08:57:38 -0400 From: tony@bioc1.biosym.com (Tony Schmidt) Message-Id: <9508041257.AA07441@east1.biosym.com> To: ccl.net!chemistry-request@bioc1.biosym.com Subject: Re: CCL:Question about Electrostatics in Molecular Mechanics Cc: tony@biosym.com, chemistry@ccl.net On Aug.8 Bill Welsh wrote: Dear Netters, Please help settle a small controversy. For a particular molecular mechanics calculation, I derived partial atomic charges from bond dipole moment measurements made in nonpolar solvent. I then ran molecular mechanics calculations using an in vacuo dielectric constant of 1.00. A colleague argues that I must use a dielectric constant appropriate to the nonpolar solvent from which the charges were derived (e.g., 2-3). It seems to me that the charges already embody the effect of the nonpolar solvent's dielectric constant, and so using the in vacuo value of 1.00 is okay and perhaps preferred. Stated another way, it seems to me that using these charges AND the dielectric of 2-3 would count the effect of the nonpolar solvent twice. Any opinions on this matter are appreciated. Thanks ... Bill Welsh Dept. of Chemistry Univ. of Missouri-St. Louis ----------------------------------------------------- If you have dipole moments in solvent, you may have to scale them for vacuum using e.g. strategy by Sharp et al, J.Phys.Chem. 96, 3822 (1992). Tony Schmidt tony@biosym.com From owner-chemistry@ccl.net Fri Aug 4 10:31:20 1995 Received: from FCRFV2.NCIFCRF.GOV for XUDONG@FCRFV2.NCIFCRF.GOV by www.ccl.net (8.6.10/930601.1506) id KAA11586; Fri, 4 Aug 1995 10:28:46 -0400 Date: Fri, 4 Aug 1995 10:29:12 -0400 (EDT) From: Dong Xu To: chemistry@ccl.net Message-Id: <950804102912.20200508@FCRFV2.NCIFCRF.GOV> Subject: Re: Question about Electrostatics in Molecular Mechanics > Dear Netters, > > Please help settle a small controversy. For a particular molecular > mechanics calculation, I derived partial atomic charges from bond > dipole moment measurements made in nonpolar solvent. I then ran > molecular mechanics calculations using an in vacuo dielectric constant > of 1.00. A colleague argues that I must use a dielectric constant > appropriate to the nonpolar solvent from which the charges were > derived (e.g., 2-3). It seems to me that the charges already embody > the effect of the nonpolar solvent's dielectric constant, and so using > the in vacuo value of 1.00 is okay and perhaps preferred. Stated another > way, it seems to me that using these charges AND the dielectric of 2-3 > would count the effect of the nonpolar solvent twice. > > Any opinions on this matter are appreciated. > > Thanks ... > > Bill Welsh > Dept. of Chemistry > Univ. of Missouri-St. Louis Dear Bill, It is still controversial how to choose the dielectric constant in molecular mechanics calculation. Some choose 1, some choose larger than 1, and some choose a distance dependent (e.g. 1/r) parameter. The current point charges were calculated under the condition without perturbation of external electric field. When two charges interact with each other, not only the effective point charges move accordingly to have the dielectric effect, but also eletrons redistribute themselves to have the so-called electronic polarizability. So if you choose a dielectric constant of 1.00, you can model dielectric effect of ion (nuclei), but you do not account completely for the polarizability of electron. That's why many people attempt to use a dielectric constant large than one. It was suggested 2.0 for proteins and 4.0 for membranes. But since there is no really good model to include the polarizability of electron, it is still acceptable to use a dielectric constant of 1.00. It seems that dynamics of biopolymers is not very sensitive to electrostatics in many cases. For example, many simulations used an 8A distance cutoff for Coulomb forces, which loses a significant portion of electrostatics, but good results were still obtained anyway. Dong Xu Laboratory of Mathematical Biology NCI/Frederick Cancer Research and Development Center P.O. Box B, Bldg. 469, #151 Frederick, MD 21702-1201 From owner-chemistry@ccl.net Fri Aug 4 10:46:28 1995 Received: from remcure.bmb.wright.edu for cletner@remcure.bmb.wright.edu by www.ccl.net (8.6.10/930601.1506) id KAA11956; Fri, 4 Aug 1995 10:40:24 -0400 Received: by remcure.bmb.wright.edu (931110.SGI/921111.SGI.AUTO) for CHEMISTRY@ccl.net id AA28374; Fri, 4 Aug 95 13:34:25 -0400 Date: Fri, 4 Aug 1995 13:13:13 -0400 (EDT) From: Charles Letner Subject: Benchmark summary To: Computational Chemistry List Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Hello all, Two weeks ago I possed the question about benchmarks. As promixed here is the summary. Thanks to all who responded! Chuck Charles Letner Wright State University Department of Biochemistry Dayton, OH 45435 e-mail: cletner@remcure.bmb.wright.edu Original post: Hello everyone, I would like to start adding some basic benchmarking to some code I have (both C/C++ and fortran). But I have a couple of question. First, what are your opinions on various benchmarks. For example, cpu time could be one benchmark. However I see this as problematic. I'm working on moving code to parallel machines. CPU time doesn't appear to me to take into account that the code could potentially take up more cpu time due to overhead involved in running multipule cpus while executing much faster in terms of "wall clock" time, ie/ user time. Mflops seems like an alternative that would overcome this but may be harder to implement. This brings me to my second question. How are you implementing these methods into your code? I'd be interested in function calls, libraries, etc... that could be used at runtime. I would like to have methods that are portable across different plateforms as the ultimate idea is to compare benchmarks of actual runs on multipule plateforms. I would also be interested in any references on this topic. I sincerely appretiate any time you take in responding. I will summarize any responses to the net towards the end of next week. Regards, Chuck Charles Letner Wright State University Department of Biochemistry Dayton, OH 45435 e-mail: cletner@remcure.bmb.wright.edu Responses: From: Hinsen Konrad Well, you should calculate whatever is relevant for you. CPU time is relevant if you are billed for it, and wall clock time is relevant because you have to wait for it. So why not have both? MFLOPS are pretty useless; they express to speed of a piece of hardware, and that with reference to hardware-dependent units. I doubt that there is anything really portable for parallel machines. But every operating system provides some functions for finding the CPU time; it should be in the documentation. IRIX, like many other Unix versions, has the system function times() to be called from C and C++ programs and the Fortran functions etime() and dtime(). I wouldn't worry too much about portability - just write a single machine-dependent function to calculate what you want; it won't be more than a couple of lines. ------------------------------------------------------------------------------- Konrad Hinsen | E-Mail: hinsenk@ere.umontreal.ca Departement de chimie | Tel.: +1-514-343-6111 ext. 3953 Universite de Montreal | Fax: +1-514-343-7586 C.P. 6128, succ. A | Deutsch/Esperanto/English/Nederlands/ Montreal (QC) H3C 3J7 | Francais (phase experimentale) ------------------------------------------------------------------------------- From: Joe Leonard Chuck - can you use time-to-solution as a benchmark? This lets you factor in things like workstation clusters and parallel machines, etc, and dispenses with individual timing on routines... If you have a set of molecules you want to perform calculations on, can you use the time it takes to carry out the calculations in various computing enviromments. This way, you can tune individual pieces of code however the vendor suggests and get a consistent measure. Might be full of it, but we use it here when measuring the efficiency of things like conformational analysis when multi-machine enviromnents are used... Joe ------------------------------------------------------------------------ Joe Leonard Wavefunction Inc. 18401 Von Karman, Suite 370 Irvine, CA 92715 I am a professional... 714-955-2120 do not attempt this at home. 714-955-2118 fax jle@wavefun.com From: "Timothy G. Mattson" Chuck, You want to use Wall time. CPU time is a poor choice since it doesn't always include time spent inside the OS on behalf of the benchmark. For example, CPU time doesn't pick up time spent managing ethernet conflicts. MFLOPS only works as a performance metric when you have a mathematical definition of the operation counts. For example, in counting flops, how many flop's do you assign to a square root. The actual count varries from one machine to another making MFLOPS a non-uniform metric. The only way to go is wall time. In terms of portable parallel computing, you have many options. A popular choice is TCGMSG. This is a small and easy to learn message passing environment designed by and for chemists. You also could go with PVM or MPI. If you want to learn more about portable parallel programming environments, take a look at my paper on the topic which you can grab by anonymous ftp export.ssd.intel.com pub/tmp/mattson/ijsa.ps This is a temporary location, so grab the paper right away if you are interested. --Tim -------------------------------------------------------------------- Timothy G. Mattson, Ph.D. - Senior Scientist Intel Supercomputer Systems Division | Phone: (503) 677-5627 Mail Stop CO1-15 | FAX: (503) 677-6324 5200 N.E. ELAM Young Pkwy | email: tgm@ssd.intel.com Hillsboro, OR 97124-6497 -------------------------------------------------------------------- From: ross@cgl.ucsf.EDU Parallel machines typically dedicate nodes to a job, so wallclock time is a 1st approximation, although traffic between nodes from other jobs could have an impact. The fortran cpu & wallclock time routines from Amber (along w/ the compiling system) are freely available for use & redistribution, if you want a wrapper to hide system dependencies. If you're interested, I'll make up a package w/ the relevant stuff. Bill Ross [The package is at] The package is in anon ftp on odin.ucsf.edu under amber/CompSys.tar.Z Bill From: Jeyapandian Kottalam On all Cray platforms we use wallclock times (call secondr or rtf). On vector-parallel platform (traditional ymp type) wallclock times are meaningful only when the machine is dedicated for the benchmark job. This is indeed how we run benchmarks, i.e., by taking the machine down and bringing it up accesible only to selected users. On massively parallel system (Cray T3D) the benchmarking job can run during batch time to get a pretty good measurement. This is because when a job starts running a certain number of processors get dedicated to that job. But the communication network is being shared with other jobs and that can make the measurement slightly inaccurate. Kottalam Phone: (USA) 612 683 3622 Cray Research, Inc. From: Matt Clark Regarding what measurement to use in a benchmark, I feel you should measure what you are really interested in. Namely "how long does it take to run", or wall clock time. If you use CPU time you would add up the time for each individual CPU, and that isn't a very useful measure of the performance increase of a parallel computer. Wall clock time takes into account any paging/swapping/io that the program causes, which are important practical issues. From: Peter Shenkin > From chemistry-request@ccl.net Fri Jul 21 14:21:18 1995 > ...Most [responses] indicate that "wall clock" time > is probably one of the better benchmarks. The most common reason: it > takes into account issues such as multipule processors, memory swapping, > etc.. To eliminate factors such as other jobs and other users one > indiviual mentioned that they make the machine unavailable to other users > while the benchmark job is running. I must disagree with this conclusion. The problem is that benchmarks run on different machines without kicking off the other users will inevitably be run with different loadings on the different machines. If machine X appears faster than machine Y, you don't know whether it's because machine Y had 10 users running big jobs whereas X only had 2 users running big jobs when the benchmarks were run. This can dramatically effect things like swapping, which increases wall-clock time greatly. CPU time is still the best benchmark, IMHO. It is sufficiently insensitive to loading to be able to be fairly run on a loaded machine. Parallel code presents additional difficulties, admittedly, but the sum of the CPU times does in fact tell you a lot, especially when compared to the CPU time when the same job is run on a single processor. -P. ***************** Atheist: An evangelical agnostic. ***************** *Peter S. Shenkin, Box 768 Havemeyer Hall, Chemistry, Columbia Univ.,* *NY, NY 10027; shenkin@columbia.edu; (212)854-5143; FAX: 678-9039* ********** World music: What Bluegrass is to a Bulgarian. *********** From: Swamy Kandadai Charlie: The issues you are raising are valid and I am sure you will get several answers. What we (in performance analysis) is really interested is in finding out what machines best perform for a special application. If the application is cpu intensive (eg. most molecular dynamics codes assuming you don't save every coordinate and velocity at every step of the molecular dynamics step), then cpu is an important characteristic in a serial run. On the otherhand, if the application involves some amount of I/O as you normally see in traditional scf calculations (writing 2-e integrals and reading them at every iteration step to form the Fock matrix), the elapsed time is important in addition to cpu time since the ratio of elapsed to cpu will tell us the I/O bandwidth on that particular machine. In parallel applications, what we are interested is fast turn around time. Everybody understands taht running parallel job on a loaded system is not the best way to run a parallel job. In parallel applications, cpu time is misleading since different cpu times indicates probalble bottelnecks in load balancing. In message-passing applications, it will also tell about how much time the application is spending on the messages to happen which in turn will tell us about the bandwidth of the medium of transfer. A single measure as cpu or the elapsed time is not going to be sufficient and the application characteristics will dictate what to look for. Thanks Swamy Kandadai From owner-chemistry@ccl.net Fri Aug 4 11:16:24 1995 Received: from wugate.wustl.edu for states@rucola.wustl.edu by www.ccl.net (8.6.10/930601.1506) id LAA12778; Fri, 4 Aug 1995 11:06:07 -0400 Received: from rucola (rucola.wustl.edu [128.252.122.177]) by wugate.wustl.edu (8.6.12/8.6.11) with ESMTP id KAA18847; Fri, 4 Aug 1995 10:06:06 -0500 Received: by rucola (950215.SGI.8.6.10/930416.SGI) id KAA07838; Fri, 4 Aug 1995 10:05:55 -0500 From: states@rucola.WUStL.EDU (David J. States) Message-Id: <9508041005.ZM7836@rucola.wustl.edu> Date: Fri, 4 Aug 1995 10:05:54 -0500 In-Reply-To: BILL WELSH "CCL:Question about Electrostatics in Molecular Mechanics" (Aug 3, 4:34pm) References: <01HTN5UE7QKI9OCUY2@SLVAXA.UMSL.EDU> X-Mailer: Z-Mail (3.1.0 22feb94 MediaMail) To: BILL WELSH , chemistry@ccl.net Subject: Re: CCL:Question about Electrostatics in Molecular Mechanics Content-Type: text/plain; charset=us-ascii Mime-Version: 1.0 On Aug 3, 4:34pm, BILL WELSH wrote: > Subject: CCL:Question about Electrostatics in Molecular Mechanics > > Dear Netters, > > Please help settle a small controversy. For a particular molecular > mechanics calculation, I derived partial atomic charges from bond > dipole moment measurements made in nonpolar solvent. I then ran > molecular mechanics calculations using an in vacuo dielectric constant > of 1.00. A colleague argues that I must use a dielectric constant > appropriate to the nonpolar solvent from which the charges were > derived (e.g., 2-3).... This is an old controversy. Most of the dielectric in a nonpolar solvent is due to electronic polarization that is not modeled by explicit molecular reorientation in an MD run. On the otherhand, it may well be incorporated implicitly into the PE function. In your example, the issue is how the bond dipole moment determination was actually done. Using charges from an electronic structure calculation or experimental procedure that gave the intrinsic molecular charge distribution, even if in a solvent cavity, to calculate a Coulomb term in the a PE function would not implicitly include solvent screening. In such case, the electronic polarization dielectric would need to be explicitly incorporated into the MD calculation. For the first solvation shell, the continuum dielectric approximation starts to breakdown. The solute and a first shell solvent molecule can be modeled as interacting in vacuo in a cavity surrounded by the continuum dielectric. There is still significant screening, but it is less than the continuum dielectric. So if you really want to do the calculation correctly, electronic polarization, in addition to molecular reorientation, must be modeled explicitly. If the data reduction from an experimental protocol actually resulted in scaling the charges by the solvent dielectric used in the measurement, then the MD calculation would actually have to multiply by the dielectric: Qi' = Qi/eps Qj' = Qj/eps Eele = eps*(Qi')*(Qj')/Rij = Qi*Qj/(Rij*eps) Dipole energy follows same scaling. David -- David J. States Institute for Biomedical Computing / Washington University in St. Louis From owner-chemistry@ccl.net Fri Aug 4 12:16:32 1995 Received: from phem3 for nauss@ucmod2.che.uc.EDU by www.ccl.net (8.6.10/930601.1506) id MAA14079; Fri, 4 Aug 1995 12:12:10 -0400 Received: from UCBEH.SAN.UC.EDU (SMTPUSER@UCBEH) by phem3.acs.ohio-state.edu (PMDF V4.2-13 #5888) id <01HTOAZ3QIW09CFPEZ@phem3.acs.ohio-state.edu>; Fri, 4 Aug 1995 12:11:57 EDT Received: from ucmod2.che.uc.edu by UCBEH.SAN.UC.EDU (PMDF V5.0-4 #7238) id <01HTOAY7R0FGHV1BQ6@UCBEH.SAN.UC.EDU> for CHEMISTRY@ccl.net; Fri, 04 Aug 1995 12:11:13 -0500 (EST) Received: by ucmod2.che.uc.edu (920330.SGI/920502.SGI.AUTO) for @uc.edu:CHEMISTRY@ccl.net id AA04753; Fri, 04 Aug 1995 12:17:03 -0400 Date: Fri, 04 Aug 1995 12:17:03 -0400 From: nauss@ucmod2.che.uc.EDU (Jeffrey L. Nauss) Subject: Re: CCL:Question about Electrostatics in Molecular Mechanics To: CHEMISTRY@ccl.net Reply-to: nauss@ucmod2.che.uc.EDU Message-id: <9508041617.AA04753@ucmod2.che.uc.EDU> Content-transfer-encoding: 7BIT References: <01HTN5UE7QKI9OCUY2@SLVAXA.UMSL.EDU> >From: states@rucola.WUStL.EDU (David J. States) >For the first solvation shell, the continuum dielectric approximation >starts to breakdown. The solute and a first shell solvent molecule can >be modeled as interacting in vacuo in a cavity surrounded by the continuum >dielectric. There is still significant screening, but it is less than the >continuum dielectric. So if you really want to do the calculation correctly, >electronic polarization, in addition to molecular reorientation, must >be modeled explicitly. The question of calculating electronic polarization explicitly may be an important one in several situations. In semi-empirical and ab initio calculations, I believe polarization is handled explicitly. (Correct me if I am wrong on this.) However, I am not familiar with any empirical potential energy function that incorporates polarization except for that developed in Wilma Olson's group for nucleic acids (Srinivasan and Olson, J. Biomol. Struc. Dyn. (1987) vol 4, pp 895-938.) They included in their non-bonded interactions an explicit charge induced dipole effect between atoms: Dij = 165.9 (delta(i)^2* alpha(j) + delta(j)^2 * alpha(i)) ---------------------------------------------------- epsilon where delta is the partial atomic charge, alpha is the atomic polarizability, and epsilon is the dielectric constant. This term accounts for stacking interactions in nucleic acids. I have never seen a similar term employed in any other forcefield and have ofter wondered why. Comments? Jeff Nauss **************************************************************************** * UU UU Jeffrey L. Nauss, PhD * * UU UU Director, Molecular Modeling Services * * UU UU Department of Chemistry * * UU UU CCCCCCC University of Cincinnati * * UU UU CCCCCCCC Cincinnati, OH 45221-0172 * * UUUU CC * * CC Telephone: 513-556-0148 Fax: 513-556-9239 * * CC * * CCCCCCCC e-mail: nauss@ucmod2.che.uc.edu * * CCCCCCC http://www.che.uc.edu/~nauss * **************************************************************************** From owner-chemistry@ccl.net Fri Aug 4 12:46:22 1995 Received: from sinfonix.rz.tu-clausthal.de for pctw@sun.rz.tu-clausthal.de by www.ccl.net (8.6.10/930601.1506) id MAA15083; Fri, 4 Aug 1995 12:43:28 -0400 Received: from gomorra.rz.tu-clausthal.de (9E+OvFcvR0EgI71QPSuZorMXxZVIPWDn@gomorra.rz.tu-clausthal.de [139.174.2.5]) by sinfonix.rz.tu-clausthal.de (8.6.12/8.6.12) with ESMTP id SAA16550 for ; Fri, 4 Aug 1995 18:42:27 +0200 Received: from remote3.slip.tu-clausthal.de (remote3.slip.tu-clausthal.de [139.174.8.19]) by gomorra.rz.tu-clausthal.de (8.6.12/8.6.12) with SMTP id SAA05773 for ; Fri, 4 Aug 1995 18:42:24 +0200 Message-Id: <199508041642.SAA05773@gomorra.rz.tu-clausthal.de> X-Sender: pctw@idefix.rz.tu-clausthal.de X-Mailer: Windows Eudora Version 1.4.4 Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Fri, 04 Aug 1995 18:41:20 +0100 To: chemistry@ccl.net From: pctw@sun.rz.tu-clausthal.de (Thomas Weber) Subject: WINNMR data format? Can anybody describe me the data format of the WINNMR program? Thomas Weber -- pctw@sun.rz.tu-clausthal.de From owner-chemistry@ccl.net Fri Aug 4 17:01:29 1995 Received: from server1.ctc.com for theochem@ctc.com by www.ccl.net (8.6.10/930601.1506) id QAA20657; Fri, 4 Aug 1995 16:59:52 -0400 Received: by server1.ctc.com (5.65/DEC-Ultrix/4.3) id AA13953; Fri, 4 Aug 1995 17:00:37 -0400 Received: by pauling.ctc.com (940816.SGI.8.6.9/930416.SGI.AUTO) for chemistry@ccl.net id RAA12978; Fri, 4 Aug 1995 17:55:37 -0400 From: theochem@ctc.com (Douglas Smith) Message-Id: <199508042155.RAA12978@pauling.ctc.com> Subject: DEC OSF/1 upgrade question To: chemistry@ccl.net (Computational Chemistry List) Date: Fri, 4 Aug 1995 17:55:31 -0500 (EDT) X-Mailer: ELM [version 2.4 PL22] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 744 I have a DEC Alpha running OSF/1 version 1.3. The graphics on this machine are the Kubota Denali graphics, running KWS 1.3.4. I wouldlike to upgrade my system to Digital UNIX (formerly OSF/1 v 3.2B). Is anyone aware of any problems this will cause me with my graphics system? Do I need to upgrade the KWS software as well (if that is even possible)? Has anyone done this and can they offer any advice? Please respond directly to me - I doubt this is of general interest to waste any more bandwidth than I am doing now. Doug -- Douglas A. Smith, Ph.D., President The DASGroup, Inc. 325 Beaver Court Johnstown, PA 15905 voice: (814) 255-6043 fax: (814) 255-6487 email: theochem@ctc.com Disclaimer: Whatever I say is my own fault. From owner-chemistry@ccl.net Fri Aug 4 17:16:29 1995 Received: from www.hyper.com for vuckovic@hyper.com by www.ccl.net (8.6.10/930601.1506) id RAA20696; Fri, 4 Aug 1995 17:00:35 -0400 Received: (from vuckovic@localhost) by www.hyper.com (8.6.9/8.6.9) id OAA00188; Fri, 4 Aug 1995 14:33:00 -0400 From: Dragan Vuckovic Message-Id: <199508041833.OAA00188@www.hyper.com> Subject: HyperChem Workshops Announcement To: chemistry@ccl.net Date: Fri, 4 Aug 1995 14:32:59 -0400 (EDT) Cc: vuckovic@www.hyper.com (Dragan Vuckovic), polowin@www.hyper.com (Joel Polowin) X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 1475 Dear CCLers, Hypercube, Inc. is pleased to announce that it is organizing a series of workshops on molecular modeling, computational chemistry, and the use of HyperChem, ChemPlus and HyperNMR. It is offering a three-day introductory (HyperChem's GUI, Molecular Mechanics, Approximate and Semi-empirical MO methods, ab initio MO methods, Molecular Dynamics, Molecular Spectroscopy and an introduction to QSAR. The course will assume some basic knowledge of chemistry and a very moderate knowledge of PC computers and Microsoft Windows) and advanced (Molecular Modeling techniques, HyperChem's GUI - Advanced Features, QSAR and QSPR, Systems with Periodic Boundaries, Molecular Dynamics and Monte Carlo Methods, Thermodynamics and Kinetics, Script Commands and Customization, and linking HyperChem to other programs - Visual Basic and Excel) courses. Workshops on site are also available, as well as one-day courses on specific issues. Please contact Hypercube, Inc. (vuckovic@hyper.com) or HyperChem dealers for more details on workshops, or visit Hypercube's home page (URL: http://www.hyper.com). ___________________________________________________________________ Dr. Dragan Lj. Vuckovic phone: (519) 725-4040 International Marketing Manager fax: (519) 725-5193 Hypercube, Inc. e-mail: vuckovic@hyper.com 419 Phillip Street URL: http://www.hyper.com Waterloo, Ontario Canada, N2L 3X2 From owner-chemistry@ccl.net Fri Aug 4 17:46:32 1995 Received: from tin for rao@ddix4.monsanto.com by www.ccl.net (8.6.10/930601.1506) id RAA21312; Fri, 4 Aug 1995 17:35:11 -0400 Received: from ddix4.monsanto.com by tin (5.65/Monsanto1.7) id AA26825; Fri, 4 Aug 1995 16:34:40 -0500 Received: by ddix4.monsanto.com id AA17849; 931110.SGI/CRI-%I%; Fri, 4 Aug 95 16:34:39 -0500 From: "Shashi Rao" Message-Id: <9508041634.ZM17847@ddix4.monsanto.com> Date: Fri, 4 Aug 1995 16:34:38 -0500 X-Mailer: Z-Mail (3.1.0 22feb94 MediaMail) To: CHEMISTRY@ccl.net Subject: Chicago computational chemistry club seminar Cc: rao@ddix4.monsanto.com Content-Type: text/plain; charset=us-ascii Mime-Version: 1.0 August 4, 1995 Chicago Computational Chemistry Club (CCCC) Dear CCCC participant, The next seminar talk under the CCCC auspicies will be held on Friday, the 18th August, 1995 at Searle R & D facilities, as outlined below. Speaker : Professor Terry P. Lybrand University of Washington Molecular Bioengineering Program Title : Strategies and Techniques for Membrane Protein Modeling Venue : Searle Research and Development, Conf. Rm. A-3B&C, 4901 Searle Parkway, Skokie IL 60077. Date : AUGUST 18, 1995. Pizza and soft drinks will be served from 6:30 - 7:30. Participants in the 210th National ACS meeting who happen to be in Chicago on the 18th August, are most welcome to attend the seminar. Please give us a call or e-mail, if you can make it (shashi rao :708-982-4545, e-mail : rao@ddix4.monsanto.com ; Charles Hutchins : hutchinsc@randb.abbott.com (708)-937-4971). If you cannot attend, please call anyway to confirm or change your address. Hope to see you again soon. Best wishes, Shashi Rao ================================================================ The Chicago Computational Chemistry Club Presents Professor Terry P. Lybrand University of Washington Molecular Bioengineering Program "Strategies and Techniques for Membrane Protein Modeling" Friday, the 18th August, 1995 at 7:30 p.m. Abstract: ======== Integral membrane receptor proteins frequently mediate signal transduction across cell membranes in both bacteria and higher organisms. As these receptors often play a key role in regulation of important physiologic processes, they can be attractive drug therapy targers. Detailed understanding of the three-dimensional structures of these receptor proteins would undoubtedly contribute greatly to a general understanding of the molecular mechanisms of signal transduction and facilitate structure-based drug design efforts, but formidable technical challenges prohibit the routine determination of high resolution structures for most membrane proteins at present. In favorable cases, molecular modeling studies can be coupled with appropriate biophysical experiments to generate plausible 3D models for some membrane receptor proteins. Projects to generate 3D structures for several bacterial and mammalian membrane receptor proteins will be presented, and some modeling techniques and experimental biophysical methods that have proven most useful in these exercises will be discussed. Directions : =========== To get to Searle, exit on Dempster (East) from I-94 (Edens expressway). Turn right (South) on Skokie Blvd (Rt. 41). After crossing the signal light at Main street, turn right (West) into Searle parkway. The visitor parking lot is on the north side of the road after crossing two railroad tracks. Enter the building at the East end and check with the security guard house. As earlier, one of us will lead the way to the seminar room. For any questions on the directions, please call shashi rao @ 708-982-4545.