From Lutz.Ehrlich@EMBL-Heidelberg.de Mon Nov 17 04:33:46 1997 Received: from stork.EMBL-Heidelberg.DE for Lutz.Ehrlich@EMBL-Heidelberg.de by www.ccl.net (8.8.3/950822.1) id EAA18186; Mon, 17 Nov 1997 04:32:15 -0500 (EST) From: Received: from cuckoo.embl-heidelberg.de (cuckoo.EMBL-Heidelberg.DE [192.54.41.16]) by stork.EMBL-Heidelberg.DE (8.8.4/8.8.4) with ESMTP id KAA02822; Mon, 17 Nov 1997 10:32:14 +0100 (MET) Received: by cuckoo.embl-heidelberg.de (8.7.1) id JAA05368; Mon, 17 Nov 1997 09:32:09 GMT Date: Mon, 17 Nov 1997 09:32:09 GMT Message-Id: <199711170932.JAA05368@cuckoo.embl-heidelberg.de> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit To: Richard Gillilan CC: chemistry@www.ccl.net Subject: Re: CCL:Rigid body dynamics In-Reply-To: <346B3D33.1CFB@tc.cornell.edu> References: <346B3D33.1CFB@tc.cornell.edu> X-Mailer: VM 6.34 under Emacs 19.34.1 Richard Gillilan wrote: > A number of programs (like XPLOR) have implemented the virtual > rigid body dynamics of T. Head-Gordon and C. Brooks > (Biopolymers vol 31, 77-100 (1991). I've seen a few notes > on CCL about "Moldy" that sound like it too can do this. I inspected Moldy's capabilities for rigid body protein docking (which I right now do with XPLOR); it also implements the Newton-Euler equations of motion which are used to propagate collections of rigid bodies. Moldy can be downloaded from earth.ox.ac.uk using anon ftp; relevant files are in the /pub directory (information taken from Moldy 2.12). The author is Keith Refson (keith@earth.ox.ac.uk). > Also: > > 1. Does anyone know of recent studies using this technique, > particularly applied to the protein-protein docking? The recent protein-protein docking studies that I know of seem to choose different ways of simplifying the proteins involved: Either a simplified energy function (which takes into account only that part of the interaction deemed relevant by the authors) is minimized using a MonteCarlo-Simulated Annealing-like protocol, or the minimum is found by complete enumeration (as in the case of the Fourier-convolution based approaches). All these approaches trade a detailed description of the dynamics for docking speed. On the other side, some MD studies of docking have been performed (cf. Di Nola et al., 1994). I haven't found any studies on protein-protein docking using Head-Gordon's VRBD. As part of my current project, I'm studied protein-protein docking using VRBD within XPLOR. Let me know if you're interested in my experiences. > 2. How does it compare with SHAKE type constraint methods: > faster/slower or just different? Just different would be a better description, as you constrain the bond angles and dihedrals as well (dependent of course, on the way you choose your rigid bodies). Second, in the SHAKE approach, each atom still experiences the force due to it's interaction with the rest of the system. In VRDB (as proposed by Head-Gordon) , a force/torque acting on a rigid body is derived from the forces acting on it's constitutive atoms. Third, Head-Gordon in her work enforces planarity of the peptide bond by turning it into a rigid body. Thus, your dynamics will differ from a plain bond-length SHAKE simulation. Hope this helps, Lutz Ehrlich --------------------- Lutz Ehrlich Structural Biology EMBL Meyerhofstr. 1 D-69012 Heidelberg Germany email: lutz.ehrlich@embl-heidelberg.de web : http://www.embl-heidelberg.de/~ehrlich phone: +49-6221-387-140 fax : +49-6221-387-517 From rabe@fu-berlin.de Mon Nov 17 08:34:17 1997 Received: from mail.chemie.fu-berlin.de for rabe@fu-berlin.de by www.ccl.net (8.8.3/950822.1) id IAA18852; Mon, 17 Nov 1997 08:00:34 -0500 (EST) Received: from Fermi.Chemie.FU-Berlin.DE(fermi.chemie.fu-berlin.de[160.45.26.7]) (1838 bytes) by mail.chemie.fu-berlin.de via smail with P:smtp/R:bind_hosts/T:inet_zone_bind_smtp (sender: ) id for ; Mon, 17 Nov 1997 14:00:31 +0100 (MET) (Smail-3.2.0.98 1997-Oct-16 #25 built 1997-Nov-07) Received: by Fermi.Chemie.FU-Berlin.DE (Smail3.2.0.98) from Fermi.Chemie.FU-Berlin.DE (127.0.0.1) with smtp id ; Mon, 17 Nov 1997 14:00:20 +0100 (MET) Sender: rabe@www.ccl.net Message-ID: <34703FE3.237C@fu-berlin.de> Date: Mon, 17 Nov 1997 14:00:19 +0100 From: Bjoern Rabenstein Organization: Freie Universitaet Berlin, Germany X-Mailer: Mozilla 3.0 (X11; I; IRIX 5.3 IP22) MIME-Version: 1.0 To: Computational Chemistry Mailing List Subject: G94: How to set parameters for CHELPG? Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Dear CCLers! I'm currently trying to do some experiments with the CHELPG routine of Gaussian94 to fit atomic partial charges to the electrostatic potential. I have problems to change or add parameters: (i) How additional van-der-Waals radii are set? e.g. radius of Fe is unknown, but my molecule contains a Fe atom. (ii) How can I change (if possible?!?) the maximum radius, up to where grid points are considered for the fit? Has anybody already done this? And what was the result? (Of course, as already discussed in this list, the monopole will dominate more and more with increasing radius, thus in general absolute values of the charges will decrease, if the molecule has over-all electroneutrality. But are their any special changings of single charges possible?) Bye! -- Bjoern Rabenstein * PhD student * Freie Universitaet Berlin Inst. f. Kristallographie * AG Knapp * Takustrasse 6 * D-14195 Berlin [phone] +49-30-838-3484 [fax] +49-30-838-3464 [email] rabe@fu-berlin.de [WWW] http://www.chemie.fu-berlin.de/~rabe/ From Mehmet.Kabak@science.ankara.edu.tr Mon Nov 17 08:45:44 1997 Received: from eros.science.ankara.edu.tr for Mehmet.Kabak@science.ankara.edu.tr by www.ccl.net (8.8.3/950822.1) id HAA18808; Mon, 17 Nov 1997 07:38:32 -0500 (EST) Received: from localhost (kabak@localhost) by eros.science.ankara.edu.tr (8.8.7/8.8.7) with SMTP id PAA12238 for ; Mon, 17 Nov 1997 15:22:22 +0200 Date: Mon, 17 Nov 1997 15:22:21 +0200 (EET) From: "Res.Ass.Mehmet Kabak" Reply-To: "Res.Ass.Mehmet Kabak" To: chemistry@www.ccl.net Subject: results of mopac6.0 In-Reply-To: <3.0.3.32.19971114124937.00688408@highdent> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear Coleagues, I am using Mopac6.0 and it gives different (small differences) results for the water molecule for the followings: input file for MOPAC6.0 --------------- AM1 GEO-OK XYZ PRECISE water molecule form O .0000000 0 .000000 0 .000000 0 0 0 0 H .9000000 1 .000000 0 .000000 0 1 0 0 H .9000000 1 109.000000 1 .000000 0 1 2 0 ------------- AM1 GEO-OK XYZ PRECISE water molecule H2 O HEAT OF FORMATION = -59.240785 KCAL ELECTRONIC ENERGY = -493.269935 EV CORE-CORE REPULSION = 144.707391 EV DIPOLE = 1.86046 DEBYE NO. OF FILLED LEVELS= 4 IONIZATION POTENTIAL= 12.464150 EV MOLECULAR WEIGHT = 18.015 SCF CALCULATIONS = 10 COMPUTATION TIME = .050 SECONDS O .0000000 0 .000000 0 .000000 0 0 0 0 -.3827 H .9612272 1 .000000 0 .000000 0 1 0 0 .1914 H .9612972 1 103.534765 1 .000000 0 1 2 0 .1914 input file for MOPAC6.0 ----------- AM1 GEO-OK PRECISE water molecule O .0000000 0 .000000 0 .000000 0 0 0 0 H .9000000 1 .000000 0 .000000 0 1 0 0 H .9000000 1 109.000000 1 .000000 0 1 2 0 --------- AM1 GEO-OK PRECISE H2 O HEAT OF FORMATION = -59.240778 KCAL ELECTRONIC ENERGY = -493.261534 EV CORE-CORE REPULSION = 144.698991 EV DIPOLE = 1.86056 DEBYE NO. OF FILLED LEVELS = 4 IONIZATION POTENTIAL = 12.464083 EV MOLECULAR WEIGHT = 18.015 SCF CALCULATIONS = 6 COMPUTATION TIME = .060 SECONDS O .0000000 0 .000000 0 .000000 0 0 0 0 -.3827 H .9613491 1 .000000 0 .000000 0 1 0 0 .1913 H .9613492 1 103.522088 1 .000000 0 1 2 0 .1913 which is the correct result and what are the differences between AM1 GEO-OK XYZ PRECISE and AM1 GEO-OK PRECISE commands? Many thanks, Mehmet Kabak science@ankara.edu.tr From stanley@stan.ch.pwr.wroc.pl Mon Nov 17 11:33:56 1997 Received: from stan.ch.pwr.wroc.pl for stanley@stan.ch.pwr.wroc.pl by www.ccl.net (8.8.3/950822.1) id KAA20110; Mon, 17 Nov 1997 10:51:07 -0500 (EST) Received: from localhost (stanley@localhost) by stan.ch.pwr.wroc.pl (8.8.5/8.8.5) with SMTP id QAA16708; Mon, 17 Nov 1997 16:47:09 +0100 Date: Mon, 17 Nov 1997 16:47:09 +0100 (MET) From: Stanislaw Kucharski To: "Douglas E. Stack" cc: "chemistry@www.ccl.net" , Stanislaw Kucharski Subject: Re: CCL:G:Transition state search and link 9999 error In-Reply-To: <01BCF2A2.EDCD4DD0@zinc.unomaha.edu> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII This is not PM-3 to be blamed. It is probably a program fault. We have similar problems with G94 running under Unix (IBM and SGI). The computation is frequently broken with a message: Error termination via Lnk1e in .... and here comes /directory/module.exe Is it a bug in G94? S.Kucharski ********************************************************************* * Stanislaw Kucharski * Tel. +48 71 3202862 * * Institute of Organic and * +48 71 3202426 * * Polymer Technology * * * Wroclaw Technical University * Fax +48 71 3203678 * * ul. Wybrzeze Wyspianskiego 27 * * * 50-370 Wroclaw, Poland * * ********************************************************************* * E-mail: stanley@stan.ch.pwr.wroc.pl * * kucharski@itots.ch.pwr.wroc.pl * ********************************************************************* On Sun, 16 Nov 1997, Douglas E. Stack wrote: > > When trying to do a transition state search using the PM3 semi-empirical = > method, the program always quits after one optimization cycle and gives = > the following message: > > GradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGradGrad > > > > THERE'S SMALL CHOICE IN A BOWL OF ROTTEN APPLES. > SHAKESPEARE > Error termination request processed by link 9999. > Error termination via Lnk1e in /disk2/gaussian/g94/l9999.exe. > Job cpu time: 0 days 0 hours 14 minutes 0.9 seconds. > File lengths (MBytes): RWF=3D 6 Int=3D 0 D2E=3D 0 Chk=3D 3 = > Scr=3D 1 > > This happens using either OPT=3DTS with one structure or OPT=3DQST2 = > using two structures. > What's happening?? =20 > Program : > Gaussian 94: SGI-G94RevB.3 30-May-1995 > 13-Nov-1997 > > > > Douglas E. Stack=09 > Assistant Professor > Department of Chemistry > University of Nebraska at Omaha > Omaha, NE 68182-0109 > (402) 554-3647 > (402) 544-3888 (fax) > destack@unomaha.edu > > > -------This is added Automatically by the Software-------- > -- Original Sender Envelope Address: destack@unomaha.edu > -- Original Sender From: Address: destack@unomaha.edu > CHEMISTRY@www.ccl.net: Everybody | CHEMISTRY-REQUEST@www.ccl.net: Coordinator > MAILSERV@www.ccl.net: HELP CHEMISTRY or HELP SEARCH | Gopher: www.ccl.net 73 > Anon. ftp: www.ccl.net | CHEMISTRY-SEARCH@www.ccl.net -- archive search > Web: http://www.ccl.net/chemistry.html > From korkin@act.sps.mot.com Mon Nov 17 13:33:42 1997 Received: from spsem02.sps.mot.com for korkin@act.sps.mot.com by www.ccl.net (8.8.3/950822.1) id MAA22730; Mon, 17 Nov 1997 12:39:22 -0500 (EST) Received: from mogate.sps.mot.com by spsem02.sps.mot.com (4.1/SMI-4.1/Email 2.1 10/25/93) id AA00820 for chemistry@www.ccl.net; Mon, 17 Nov 97 10:39:22 MST Received: from emailsps.sps.mot.com by mogate.sps.mot.com (4.1/SMI-4.1/Email-2.0) id AA07689 for chemistry@www.ccl.net; Mon, 17 Nov 97 10:39:21 MST Received: from merchant.sps.mot.com (merchant) by act with ESMTP (1.37.109.16/16.2) id AA133918359; Mon, 17 Nov 1997 10:39:19 -0700 From: Anatoli Korkin Received: (from korkin@localhost) by merchant.sps.mot.com (8.7.1/8.7.1) id KAA14083; Mon, 17 Nov 1997 10:39:19 -0700 (MST) Date: Mon, 17 Nov 1997 10:39:19 -0700 (MST) Message-Id: <199711171739.KAA14083@merchant.sps.mot.com> To: chemistry@www.ccl.net Cc: korkin@act.sps.mot.com Subject: pseudo-potentials for lanthanides X-Mailer: Siren Mail (Motif 2.02 95/05/15) Mime-Version: 1.0 Content-Id: <304_13972_879788358_1@merchant> Content-Type: text/plain; charset="us-ascii" Dear CCLers, Can you advise me regarding pseudo-potentials for lanthanides? Thermochemistry of compoundes, which include lanthanides is of the primary interest, but other properties are of interest too. Thank you in advance, Anatoli ****************************************************************** Anatoli Korkin Ph.D korkin@act.sps.mot.com Computational Chemist Tel: (602) 655-3171 Fax: (602) 655-2285 Predictive Engineering Lab Semiconductor Products Sector Arizona Technology Motorola Inc., MD M350 Laboratories 2200 West Broadway Road Mesa AZ 85202 ******************************************************************* From lavelle@mbi.ucla.edu Mon Nov 17 18:33:43 1997 Received: from rho.ben2.ucla.edu for lavelle@mbi.ucla.edu by www.ccl.net (8.8.3/950822.1) id RAA24364; Mon, 17 Nov 1997 17:40:12 -0500 (EST) Received: from ewald.mbi.ucla.edu (ewald.mbi.ucla.edu [128.97.39.21]) by rho.ben2.ucla.edu (8.8.5/8.8.5) with SMTP id NAA85200; Mon, 17 Nov 1997 13:48:35 -0800 Received: from red5.mbi.ucla.edu by ewald.mbi.ucla.edu (5.65v3.2/1.1.10.5/09Oct97-1217PM) id AA14297; Mon, 17 Nov 1997 13:48:32 -0800 Message-Id: <3.0.32.19971117135321.00a38e10@mbi.ucla.edu> X-Sender: lavelle@mbi.ucla.edu X-Mailer: Windows Eudora Pro Version 3.0 (32) Date: Mon, 17 Nov 1997 13:53:22 -0800 To: chemistry@www.ccl.net, hyperchem@hyper.com From: Laurence Lavelle Subject: Replies received: Improving ab initio computational speed Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Thanks for all the replies, they are listed below. (It appears that there was a 4 day delay from the time I posted my message to CCL.) My original question was: >Date: Sat, 08 Nov 1997 19:05:12 -0800 >To: hyperchem@hyper.com, chemistry@www.ccl.net >From: Laurence Lavelle >Subject: Improving ab initio computational speed > >Two possible ways to speed up ab initio calculations are: > >1) Use a bases set with fewer bases functions first. Then, after convergence, use a bases set with more bases functions. Repeat this process until convergence is obtained with the (largest) bases set that was originally intended. > >2) Use the largest bases set with a higher 2e integral cuttoff point (e.g. 10^-7 Hartree) and after convergence decrease the 2e integral cuttoff point. Repeat until convergence is obtained with the originally intended 2e integral cuttoff point (e.g. 10^-11 Hartree). > >Do either of these approaches have potential downfalls. Is one approach better than the other? > >Thanks >Laurence Lavelle **************************************************************************** ****************** Date: Mon, 17 Nov 1997 14:03 -0600 From: "David Maxwell" To: lavelle@mbi.ucla.edu Subject: Re: Improving ab initio computational speed Dear Laurence, Both of the methods you listed may be utilized to achieve your goal of speeding up the calculations. When utilizing a lower basis set, be aware that there may not be a one-to-one correspondence of minima between different basis sets. With regards to the integral accuracy, I have only seen rare occasions where this became a problem and it was usually when I was working with elements further down on the periodic table. Something you may also want to try is optimizing with some geometrical values held fixed and then allowing them to relax in a follow-up calculation. Sincerely, Dave **************************************************************************** **************** Date: Fri, 14 Nov 97 09:47:22 -0330 From: jamesx@bohr.chem.mun.ca (James Xidos) To: LAVELLE@MBI.UCLA.EDU Subject: Geometry Optimisation The rate of convergence in a geometry optimisation is directly dependent on the quality of your initial guess, and the quality of your Hessian (force constant matrix). The best route, through experience, of optimising structures at higher basis sets is to first, optimise at a lower basis set, followed by evaluating the Hessian (force constant matrix) at that lower basis set, then input both geometry and Hessian into your higher basis set job. Your second suggestion could help speed up the calculation of the initial steps in your higher level geometry optimisation, but a good geometry and Hessian will eliminate steps. I would be more inclined to increase the SCF convergence criteria for the initial steps as opposed to manipulate the 2e integral cutoffs. James Xidos Memorial University of Newfoundland jamesx@bohr.chem.mun.ca **************************************************************************** **************** From: "Chris Murray" Date: Mon, 17 Nov 1997 17:17:45 +0000 To: lavelle@mbi.ucla.edu Subject: Improving ab initio computational speed Hi Laurence, I'm surprised you got no responses since I always imagine there is someone ready to answer this type of question. 1) Use a basis set with fewer bases functions first. You can do this but it involve some algebra unless the large basis set is just an augmented form of the first. I seem to remember the results are disappointing in going from STO-3G to 3-21G say. It did not seem to be better than forming a density matrix from a superposition of results from pre-stored isotropic density matrices for atoms in the correct basis set. Good references are: Almlof's original Direct SCF paper or Cremer and Gauss, J. Comp. Chem. Vol. 7, pp274 (1986). 2) My feeling about the 2e cutoff's is that it won't work well enough to justify doing it. Also if you are doing a Direct SCF and using a change in density matrix elements in the integral test then there is a danger of inhibiting converence - in such cases it is advisable to increase the tolerances slightly during the run. Hope that this is some help. chris **************************************************************************** **************** From: "Stavrev, Krassimir" To: "'Laurence Lavelle'" Cc: "'hyperchem@hyper.com'" Subject: RE: Improving ab initio computational speed Date: Wed, 12 Nov 1997 14:44:22 -0500 Sender: listadmin@hyper.com Reply-To: "Stavrev, Krassimir" Laurence: The first approach makes more sense to me; the second one may affect the accuracy of the calculations unpredictably. In any case, with HyperChem 5 one can only export the eigenvector matrix and reproduce properties when the matrix is imported back, but cannot use the ext-file to form an initial guess of the Fock matrix. Accordingly, the SCF information that is stored at certain point is lost when one proceeds with another single-point calculation. It would be nice to have such read-from-file initial guess option in the program; it would be especially useful for people who run lengthy calculations and want to restart the job from the point where the SCF procedure was terminated. Regards, Krassimir --- Krassimir Stavrev, PhD, Director of Scientific Support Hypercube, Inc. Florida Science and Technology Park 1115 N.W. 4th Street Gainesville, Florida 32601 Voice:(352) 371-7744/(800) 960-1871 Fax:(352) 371-3662 **************************************************************************** **************** Sender: turner@ocisgi28.unizh.ch Date: Mon, 17 Nov 1997 09:03:08 +0100 From: Alexander J Turner Organization: OCI - University of Zurich To: Laurence Lavelle Subject: Re: CCL:Improving ab initio computational speed References: <3.0.32.19971108190512.00a2a3d0@mbi.ucla.edu> Hi! In reply to your question. Both methods have merit, but there are potential problems. The main problem with the basis function idea is that it assumes that a higher basis set calculation will converge faster from your low basis set optimized structure than from your original guess. This might well not be the case. It will tend to be the case much more if you compute an exact hessian at the low basis set end point and use that as the guess hessian for the higher basis set calculations. This can be especially so for HF -> MP2/DFT work. Also, you assume that the lower basis set will be able to model your compound at all. STO3G can be so far from 6-31g* (say) that your initial STO3G calculations 'lock' you into an unrealistic part of the potential energy surface. Such considerations are much more a problem when you are looking for transition states. The second method has a problem, it is virtually of no interest for conventional SCF. The time is taken computing the 2e integrals, not converging the SCF. For direct schemes, it is of interest, but only if you are going to try and get a very tight convergence of the SCF. I.E. so tight you encounter slight convergence problems. The normal property of convergence is that the last bit takes less time than the initial stages (because DIIS has more information to work with). Thus you gain little from the lower convergence, and that little you loose from the extra optimization cycles. So - my recommendation is inheritance of the structure _and_ hessian from a lower level of theory (or basis set) with extreme caution when applied to saddle points. Best wishes Alex **************************************************************************** **************** From: Eugene Fleischmann Organization: Q-Chem, Inc. To: Laurence Lavelle Cc: hyperchem@hyper.com Subject: Re: Improving ab initio computational speed Date: Thu, 13 Nov 1997 11:14:45 -0500 Unless your starting structure is really bad, I cannot recommend approach 1) above, as it is more efficient to start with the final basis set. The optimized structure and hessian from a run with a smaller basis set may be no better than a good initial guess at the initial structure on the original structure. Nevertheless, Q-Chem can handle such projections of a smaller basis onto a larger one. Regarding approach 2) above, it is more efficient to adjust the integral cutoff iteratively within the SCF scheme rather than stopping, reading, and restarting the job using multiple single point calculations. Such a variable cutoff approach is already being used by a number of programs, including Q-Chem. Best Wishes, Eugene Fleischmann **************************************************************************** ***************** From leclerc@tammy.harvard.edu Mon Nov 17 20:33:46 1997 Received: from tammy.harvard.edu for leclerc@tammy.harvard.edu by www.ccl.net (8.8.3/950822.1) id UAA25089; Mon, 17 Nov 1997 20:08:46 -0500 (EST) Received: (from leclerc@localhost) by tammy.harvard.edu (8.8.2/8.8.2) id UAA26360; Mon, 17 Nov 1997 20:14:52 -0500 (EST) Date: Mon, 17 Nov 1997 20:14:52 -0500 (EST) From: Fabrice Leclerc Message-Id: <199711180114.UAA26360@tammy.harvard.edu> To: CHEMISTRY@www.ccl.net Subject: neamine coord Anybody could provide the X-ray coordinates of neamine or any related aminoglycoside (ribostamycin, neomycin) ? Fabrice Leclerc.