From SHAUN%JASON.DECNET@relay.the.net Mon Oct 1 18:56:28 1992 Date: 02 Oct 1992 00:56:28 -0600 (CST) From: "Shaun D. Black" Subject: PDB numbers To: chemistry@ccl.net To Bill Welch and others interested, the pertinent numbers for the Protein Data Bank (PDB) are: (516)282-3629 (Phone) (516)282-5751 (FAX) pdb@bnlchm.bitnet (BITNET) pdb@chm.chm.bnl.gov (Internet) Cheers. =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= = Shaun D. Black, PhD | Internet: shaun%jason.decnet@relay.the.net = = University of Texas | Bitnet: shaun%jason.decnet@thenic.bitnet = = Health Center at Tyler | Phone: (903)877-2806 FAX: (903)877-7558 = - Tyler, TX 75710-2003 | 8-) (Start every day with a smile...) = =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= From gobbi@sg1501.chemie.uni-marburg.de Fri Oct 2 08:49:34 1992 From: Alberto Gobbi Subject: Re: PC problem alert To: chemistry@ccl.net (Quantum Chem list) Date: Fri, 2 Oct 92 9:15:06 MET Conserning the nmerical instabilities on 286 PC i have made the following expirence: The numerical Coprozessor Intel 287 makes grates deviation if numbers are used in calculation which exceed the defined values. There is not allways a Floating point eception in these casese. So if you have this instabilities on a PC 286 it may bee that you have to: 1) change your algorithm so that all values are not outside the allowed Floa- ingpoint limits. 2) swith off your Coproz (slow!) 3) use a nother coproz (I only tried intel) 4) use a newer Coproz ( 386 PC with 387 may work, I tested 486 Proz) Alberto Gobbi From hooft@xray.chem.ruu.nl Fri Oct 2 08:54:16 1992 From: hooft@xray.chem.ruu.nl (R.W.W. Hooft) Subject: Re: PC problem alert To: chemistry@ccl.net Date: Fri, 2 Oct 92 9:26:00 MET > We have found mumerical instabilities for matrix diagonalizations on certain > PC's. The same Microsoft Fortran 5.1 double precision executable module was > used on all PC's; Sun Fortran was used under UNIX on the SPARC station. > Using a 111x111 relatively sparse matrix and a Jacobi rotation routine, > a Compuadd 286 PC produced inaccurate eigenvalues. However, a Gateway 386 PC, > a SPARC station IPC (UNIX) and the SPARC station running Microsoft Fortran > under Soft-PC all produced correct results. For this test the Gateway 386 was > stable up to the maximum size tested, 150x150. The SPARC station has shown no > instability up to 550x550. I think this is highly unlikely. As you might know, the 286 chip itself is not capable of doing floating point computations. So: If you have a program that uses floating point computations, these will be emulated (using integer computations) by the compiler. Then it should be obvious that all excessive rounding errors are caused by incorrect functioning of the emulation in the compiler. Your effect could be caused by: 1) differences in the matrices used. It is for instance known that no matrix inversion whatsoever is able to correctly invert a Hilbert matrix of, say, order 20. 2) Presence of a numerical coprocessor in one of the machines. I really don't know soft-PC, but it is certainly possible that it simulates a co-processor as well. 3) Inadvertent use of double precision, or double precision intermediate results, kept on the stack of a numerical coprocessor. 4) Differences in the compiler. It is possible that even exactly the same compiler produces different code on 286 and 386 architectures. optimization could lead to (again) double precision intermediate results or a different order of evaluation. If you can rule out these four possible causes for your result, I'll be very interested to hear more of this. -- Rob W.W. Hooft, Department of crystal and structural chemistry Bijvoet Center for Biomolecular Research, University of Utrecht The Netherlands ===== hooft@chem.ruu.nl (hooft@hutruu54.bitnet) From arne@mango.mef.ki.se Fri Oct 2 11:45:14 1992 Date: Fri, 2 Oct 92 10:45:14 +0100 From: arne@mango.mef.ki.se (Arne Elofsson) To: CHEMISTRY@oscsunb.ccl.net Subject: PEPTIDE conformations. I do not remember where Sheraga wrote that but I would like to make you a few ither suggestions: 1. If it is a small peptide (<20 aa) I think just simulating at a high temperature will sample enough. 2. I think it is extremely important that you use a good solvent model. Water should not be any problems and probably neither ethanol methanol but mor complex solvents are not rigourously tested. arne@mango.mef.ki.se From toomas@chem.ut.ee Fri Oct 2 10:49:35 1992 From: toomas@chem.ut.ee (Toomas Tamm) Subject: Re: PC problem alert To: chemistry@ccl.net Date: Fri, 2 Oct 92 11:59:44 EET > Subject: Numerical instabilities on PC's > > We have found mumerical instabilities for matrix diagonalizations on certain > PC's. The same Microsoft Fortran 5.1 double precision executable module was > used on all PC's; Sun Fortran was used under UNIX on the SPARC station. [ ... rest of article deleted ... ] The whole article does not mention wether the computers were equipped with the 80*87 math co-processors or not. If not, the compiled executable must have used its linked-in floating-point emulation software, which may not accurately reflect the behaviour of the '87 chip. I have never used MSF 5.1, but I recall >from using MSF 4.1 that there was a library that detected the presence of *87 and used its emulation if the chip was absent. It is also worth notice, that different flavors of 387 chips exist, differing in speed and also accuracy. Intel, Weitek, and Cyrix are some names that come to mind. I am not sure if the same holds for 287 chips. I have a Cyrix demo that actually shows some differences when run on different computers. Different computers should be considered 'equal' in the sense of this kind of comparison only if they all use the same chips. Only if none of the computers had a *87 chip, the 286 should be considered too different from the 386. > >From these tests we conclude that MS Fortran itself is OK. The problem might > be specific to certain Compuadd machines but I have heard related complaints > about rounding errors from colleagues so I suspect that it may be general to > the Intel 286 chip. The floating point emulation is performed using integer arithmethic, and I seriously doubt that the chips that have been in use so long time still have integer arithmetic bugs in their microcode. -- Toomas Tamm Internet : toomas@chem.ut.ee Tartu University, Dept. of Chemistry fax : 7+01434+35440 2 Jakobi st. EE-2400, Tartu, Estonia voice : 7+01434+35224 From mallon@rzmain.rz.uni-ulm.de Fri Oct 2 14:12:46 1992 Date: Fri, 2 Oct 92 13:12:46 +0100 From: mallon@rzmain.rz.uni-ulm.de To: "chemistry@ccl.net"@netway.rz.uni-ulm.de Subject: Re : PC problem alert In a message received on 2 Oct 1992, 12:45 chemistry-request@ccl.net wrote to chemistry@ccl.net (Quantum Chem list) >The numerical Coprozessor Intel 287 makes grates deviation if numbers are >used in calculation which exceed the defined values. There is not allways >a Floating point eception in these casese. So if you have this instabilities >on a PC 286 it may bee that you have to: >1) change your algorithm so that all values are not outside the allowed Floa- >ingpoint limits. >2) swith off your Coproz (slow!) >3) use a nother coproz (I only tried intel) >4) use a newer Coproz ( 386 PC with 387 may work, I tested 486 Proz) We ran into similiar trouble with a 287 intel coprozessor. Things worked fine when we switched to a 386/387 (intel )configuration. To point three of the above article : we made our 286 work fine with an IIT - 287 coprozessor. Greetings, RaMa ----------------------------------------------------------------------- | Rainer "RaMa" Mallon | | | University of Ulm, Germany | I am an instant swimmer ... | | Department of Applied Physics | ... just add water ! | | mallon@main01.rz.uni-ulm.de | | ----------------------------------------------------------------------- From AHOLDER@VAX1.UMKC.EDU Tue Oct 2 02:32:51 1992 Date: 02 Oct 1992 07:32:51 -0500 (CDT) From: Andy Holder Subject: Evans and Sutherland To: CHEMISTRY@ccl.net Anyone out there in net land know of a chemistry contact at Evans and Sutherland? Working one's way in from the cor- porate direction is icky. Andy =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= DR. ANDREW HOLDER Assistant Professor of Computational/Organic Chemistry Department of Chemistry || BITNET Addr: AHOLDER@UMKCVAX1 University of Missouri - Kansas City || Internet Addr: aholder@vax1.umkc.edu Spencer Chemistry, Room 315 || Phone Number: (816) 235-2293 Kansas City, Missouri 64110 || FAX Number: (816) 235-1717 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= From cramer@chemsun.chem.umn.edu Tue Oct 2 06:06:12 1992 Date: 02 Oct 1992 11:06:12 -0500 (CDT) From: Christopher Cramer Subject: An EO issue of sorts To: chemistry@ccl.net Fellow netters, At the considerable risk of being chastised for political correctness, I've decided to forward this to the net. We are all interested in nurturing our field and its future protagonists, so I hope many of you will agree that this is not altogether trivial. (Besides, 800 numbers are so stylish just now!) > > > The Mattel Toy Company just > > came out with a talking barbie doll - TEEN TALK BARBIE. > > Talking barbie says only four things, AND one of the things > > they chose for her to say is "MATH CLASS IS HARD". > > > > This is really unfortunate. Young kids, especially girls, > > need encouragement in math not negative reinforcement. > > > > If this bothers you, we encourage you to call the toll free > > number below and place your comments: > > > > 1-800-421-2887 Mattel's Talking Barbie > > -- Christopher J. Cramer University of Minnesota Department of Chemistry 207 Pleasant St. SE Minneapolis, MN 55455-0431 From tudor@wucmd.wustl.edu Fri Oct 2 06:56:22 1992 Date: Fri, 2 Oct 92 11:56:22 -0500 From: tudor@wucmd.wustl.edu (Tudor Oprea) To: CHEMISTRY@ccl.net Subject: RE: numerical instabilities on PCs In reply to the PCs numerical processor problem: we tried to compile MOPAC 4.0 back in Timisoara starting with the version of Dr. Kapsomenos (Thessaloniki, Greece), called Mopac 4.02, written for OS/2, with MSF 5.10 on a DOS-based 486, but it did not work. Our problem was solvable by rewriting the whole clock and i/o codes and recompiling the whole with the Lahey F77 program. I also found out that a shareware Mopac 6.0 for the PC is compiled using the same Lahey. For me, this one looks as the serious Fortran 77 compiler available for PCs (this impression is shared by friends of mine). Yet, the running file didn't reproduce the manual files. We found out that extending from 4 MB RAM to 16 MB RAM was the trick to make it work fine. Thus, what I know is that when running large matrices, the PC stores on the disk IF it has not enough RAM. Yet, our Mopac worked well when much larger RAM was used. What I suggest is: could the amount of RAM used by David Perry's team in their test be a factor? ================================================================================ = Tudor Ionel OPREA = Tel. (1-314) 935 4672 = = Research Associate = Fax. (1-314) 935 4979 = = Washington University Center for Molecular Design = Email: = = Lopata Hall, Box 1099, One Brookings Drive = tudor@wucmd.wustl.edu = = St.Louis MO 63130 = = ================================================================================ From SHAUN%JASON.DECNET@relay.the.net Tue Oct 2 07:12:42 1992 Date: 02 Oct 1992 13:12:42 -0600 (CST) From: "Shaun D. Black" Subject: Ribbons v2.0 To: chemistry@ccl.net Dear CompChem Netters: We have been using Mike Carson's elegant molecular rendering software package 'Ribbons' (v2.0) for about 6 months. It comes highly recommended for small and macromolecule rendering on the Iris workstation. But, we have one problem with it. If the .PDB input file has non-contiguous residue numbers, the program cannot read the file. We recently ran across this situation with trypsin in which the file has a few cases where residues go >from 284-285-285A-286. Ribbons gets hung up on the 'A' numbering 'break'. Has anyone written a utility to renumber sequences so as to make them acceptable to Ribbons? Thanks for your help. Cheers. =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= = Shaun D. Black, PhD | Internet: shaun%jason.decnet@relay.the.net = = University of Texas | Bitnet: shaun%jason.decnet@thenic.bitnet = = Health Center at Tyler | Phone: (903)877-2806 FAX: (903)877-7558 = - Tyler, TX 75710-2003 | 8-) (Start every day with a smile...) = =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= From troyer@heffalump.mmwb.ucsf.EDU Fri Oct 2 04:37:33 1992 To: chemistry@ccl.net Subject: Re: Protein Phrenology Date: Fri, 02 Oct 92 11:37:33 -0700 From: troyer@heffalump.mmwb.ucsf.EDU > National Center for Supercomputing Applications (NCSA) has recently released > a new geometric software with 3D graphic user interface, the tool is called > Alvis (Alpha Shape Visualizer). > One can use Alvis to construct and compute 3D structures of molecules > (proteins). The input to Alvis is any scattered point set (x, y, z > coordinates in 3D sapce), and one can render it's disired shape, or > output it's solid or surface mesh. Just a comment on this program and alpha shapes. You can think of alpha shapes as the equivalent of Connoly's Molecular Surface. A probe cuts out the inside and outside of a set of points. The applicability for a raw set of protein atoms is limited, however. The resulting surface looks very strange, since the point atoms, and not a radius surrounding the points, are used to define the shape. Also, the shapes are linearized into polygonal faces. These look quite distracting compared to the dot surfaces we are used to looking at. I did not find it useful to visualize proteins in this manner. One could theoretically use the set of points comprising the molecular or van der Waals surface as input to Alvis. But the program has a limitation of ~10,000 points, and all of the protein surfaces I happened to have lying around are larger than that. The program itself, however, is impressive. The probe size can be interactively modified, allowing one to screen for pockets or clefts on a protein surface. You can display various quantities as a function of probe size, such as the total volume or surface area. I am interesting in calculating some protein volumes, and so this feature will be useful to me. For just visualizing molecular surfaces with more than a dot surface, I would instead recommend Polyview3.0, also available from ftp.ncsa.uiuc.edu, or the SGI Explorer software. Note that all of these programs run only on SGI machines (at least until OpenGL is a reality). I am in no way associated with NCSA or SGI. I just ftp things from them occasionally. John Troyer troyer@cgl.ucsf.edu Department of Pharmaceutical Chemistry Univeristy of California San Francisco, CA 94143-0446 From nauss@wrair-emh1.army.mil Tue Oct 2 11:52:00 1992 Date: 2 Oct 92 16:52:00 EST From: nauss@wrair-emh1.army.mil Subject: Applying constraints on improper dihedrals in CHARMM To: "chemistry" Good morning, netters - I am having a problem implementing a MD technique and I am looking for some help in how to proceed. Recently there have been several papers by Boryeu Mao at Upjohn regarding mass weighted molecular dynamics. This technique is used to perform a conformational search through the space of proper dihedral angles for peptides or protein fragments. It involves weighting the atomic masses of individual atoms by some arbitrary factor to blast through potential energy barriers in the course of the MD simulation. References for the technique are listed below: 1. Mao, B. & Friedman, A. R. Biophys. J. 58, 803-805 (1990). 2. Mao, B. Biophys. J. 60, 611-622 (1991). 3. Mao, B., Maggiora, G. M. & Chou, K. C. Biopolymers 31, 1077-1086 (1991). 4. Mao, B. Biophys. J. 60, 966-973 (1991). Now to my question: In implementing the technique, one needs to apply harmonic constraints on all internal coordinates except for proper dihedral angles. That's the problem for me. I know that the CONS IC command can be used for the bond lengths and bond angles but what of the improper dihedrals? Does anyone have a suggestion for applying constraints on just the improper dihedral angles? I have already considered resetting the force constant individually for each angle in my structure. But that could be rather labor intensive unless wild cards can be used. Thanks in advance... Jeffrey Nauss -------------------------------------------------------------------------------- E-mail address: nauss@wrair-emh1.army.mil Mailing Address: Department of Gastroenterology Division of Medicine Walter Reed Army Institute of Research ATTN: MAJ Jeffrey L. Nauss, Ph.D. Washington, D.C. 20307-5100 Telephone: 202-576-3485 From rsjuds@snll-arpagw.llnl.gov Fri Oct 2 07:26:34 1992 Date: Fri, 2 Oct 92 14:26:34 -0700 From: rsjuds@snll-arpagw.llnl.gov (judson richard s) To: chemistry@ccl.net Subject: SETTLE via email??? Dear netters, Does anyone have email addresses of S. Miyamoto or P. Kollman? I'm trying to get a copy of their SETTLE program via email. ----------------------------------------------- | Richard Judson | | Center for Computational Engineering | | Sandia National Laboratories | | Livermore, CA 94551-0969 | | (510)294-1438 | | email: rsjuds@sandia.llnl.gov | | | ----------------------------------------------- From ross@cgl.ucsf.EDU Fri Oct 2 09:34:21 1992 Date: Fri, 2 Oct 92 16:34:21 -0700 From: ross@cgl.ucsf.edu (Bill Ross) To: chemistry@ccl.net Subject: for the record To correct two minor misstatements in recent weeks: 1. Charles Foley named Dave Pearlman as a major author of Amber 3.0 Rev A; in fact, Dave is a major author of Amber 4.0 and George Seibel is the author of the "Rev A" of Amber 3.0. 2. The first of the excoriated Biostructure S.A. postings claimed they are "working with ... Peter Kollman" - the actual nature of this relationship is that they are licensed to sell Amber and interface their own program with it. No scientific collaboration is involved. Bill Ross PS - Heard on NPR: the FBI is requesting a law or regulation forcing computer networks to provide hooks for tapping, with $10K penalties for (designing? implementing? installing?) networks - including local ones - without such hooks. OMB has given them the go-ahead to try to sell it on Capitol Hill.