From dlim;at;minerva.cis.yale.edu Fri Apr 8 02:55:48 1994 Received: from minerva.cis.yale.edu for dlim at.at minerva.cis.yale.edu by www.ccl.net (8.6.4/930601.1506) id CAA27536; Fri, 8 Apr 1994 02:17:02 -0400 Received: by minerva.cis.yale.edu (5.0/SMI-SVR4) id AA20663; Fri, 8 Apr 1994 02:17:16 +0500 From: dlim : at : minerva.cis.yale.edu (Dongchul Lim) Message-Id: <9404080617.AA20663(-(at)-)minerva.cis.yale.edu> Subject: normal coordinate calculation To: chemistry;at;ccl.net Date: Fri, 8 Apr 1994 02:17:15 -0400 (EDT) X-Mailer: ELM [version 2.4 PL23] Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 14481 A number of people asked me to post a summary on normal coordinate calculations. Many chemists are interested in refinement of force constants from experimental frequencies and as well as prediction of frequencies from their force field parameters. Although the basic principle of calculating normal frequencies and coordinates have not changed much since 1960's, the efficiency of computation and user-friendliness became the major concern. A number of different algorithms for the frequency calculation appear in the literature. I'm not going to judge which one is better than the other (besides, I haven't tried all of them). A review which will appear in the April issue of Current Opinion in Structural Biology by Dr. David Case (see below) may cover this issue (?). I simply compiled the replies after some trimming. The order of the replies is absolutely based on their arrival time. I'd like to thank all who kindly replied to my question. I'll try to repost the summary if I get more information. -Dongchul Lim ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==* From Bill Ross (ross(-(at)-)cgl.ucsf.edu) ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==* Amber contains a normal mode program. A blurb on the package as a whole is appended, followed by the beginning of the relevant section of the manual. Bill Ross AMBER 4.0 Molecular mechanics simulation programs, including source code and demos. Computer Req'ts: Fortran compiler, 25+ Mbytes disk Amber is a suite of programs for performing a variety of molecular mechanics based simulations on machines ranging from workstations to supercomputers. It is designed primarily for proteins and nucleic acids. A graphics front end is under development and should be available in spring of 1994. The 4.0 release has important new features. Free Energy: dynamically modified windows, potential of mean force, bond correction, thermodynamic integration and others. Normal modes: Langevin modes, finding transition states, new analysis tools. A new NMR-oriented energy minimization/dynamics program allows time-dependent and time-averaged constraints (e.g. simulated annealing) and direct fitting to NOESY or chemical shift data. The 'vanilla' energy minimization and dynamics program includes polarizability as an option. Distribution is in source code format, and a suite of demos is provided. The main release is in Unix and VMS; VM/CMS/MVS will have fewer demos; and the mostly standard Fortran source code can be ported elsewhere. Price: Academic & Development: $200 Industrial use: $15,000 Support: No guarantees of support. Email address & phone number provided. NMODE module Page 1 NMODE This program performs molecular mechanics calculations on proteins and nucleic acids, using first and second derivative information to find local minima, transition states, and to perform vibrational analyses. It is designed to read the prmtop and inpcrd files from the Amber suite of programs. There are accompanying programs nmanal (normal mode analysis) and lmanal (Langevin mode analysis) that use the output of these programs to compute molecular fluctuations and time correlation functions. Nmode was originally written at the University of California, Davis, by D.T. Nguyen and D.A. Case, based in part on code in the Amber 2.0 package. Major revi- sions were made at the Research Institute of Scripps Clinic by J. Kottalam and D.A. Case. M. Pique has provided valuable advice and help in porting it to many different machines. References. The second derivative routines are based on expressions used in the Consistent Force Field programs;[1] similar information is given by K.J. Miller, et al.,[2] although these expressions were not actually used in writing this code. The code also contains routines to search for tran- sition state, starting (generally) from a minimum. This pro- cedure uses a modification of the procedure of Cerjan and Miller[3] as described elsewhere.[4]` Langevin modes are analo- gous to normal modes, but in the presence of a viscous coupling to a continuum solvent. The basic ideas are presented by Lamm and Szabo,[5] and were implemented in the Amber environment by us.[6] General description: This program performs five tasks, depending on the value of the input variable ntrun (see below): ____________________ [1]S.R. Niketic and K. Rasmussen, The Consistent Force Field: A Documentation, Springer-Verlag, 1977. [2]R.J. Hinde and J. Anderson, J. Comput. Chem. 1989, 10, 63. [3]C. Cerjan and W.H. Miller, J. Chem. Phys. 1981, 75, 2800. [4]D. T. Nguyen and D. A. Case, J. Phys. Chem., 1985, 89, 4020. [5]G. Lamm and A. Szabo, J. Chem. Phys. 1986, 85, 7334. [6] J. Kottalam and D.A. Case, Biopolymers 1990, 29, 1409- 1421. _________________ (1) Perform a normal mode analysis from starting coordi- nates. Requires an input structure that has already been minimized, from process (4), below, or by some other method. In addition to the computation of normal mode frequencies, thermodynamic parameters are calcu- lated. (2) Search for transition state, starting (generally) from a minimum. See the references above for a detailed description of the method. (3) Perform a conjugate gradient minimization from the starting coordinates. This routine uses an IMSL library routine for this purpose, which is not supplied with this program. Persons who do not have access to the IMSL library should probably use the AMBER "min" program to carry out conjugate gradient minimizations. (Compile min in the double precision version for best conver- gence.) (4) Does a Newton-Raphson minimization from starting coordi- nates. A constant (tlamba) is added to the diagonal elements of the Hessian matrix to make it positive definite. Tlamba is chosen in a manner such that the step is always downhill in all directions. Whenever the change in energy is > emx or the rms of step length is > smx, the step length is scaled back repeatedly until the above two conditions are satisfied. Note that this rou- tine will not converge to a transition state. (5) Perform a langevin mode calculation, starting from a minimized structure. This option is similar to (1), but includes the viscous effects of a solvent in the calcu- lation. ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==* From: "Don Gregory" ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==* CHARMm has quite a good set of routines for calculating normal modes, vibrational spectra, and force-field parameter optimization of foce-constants to match the experimental IR. ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==* From: Tom Sundius U of Helsinki ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==* I am the author of a program for force field calculations (which of course also produces normal coordinates). It is called MOLVIB and has been included in the QCPE collection (#604, see QCPE Bull. Vol 11, Nr. 3). I include the description, which also gives some references. By the way, Gwinn's program does also use cartesian coordinates in the calculations, but had some problems with the calculation of the B-matrix, which I think were corrected by the Norwegian electron diffraction group. The classical method (so-called GF-method) is used in the Schachtschneider program, which was developed in the 60's. Shimanouchi and his coworkers in Japan developed a new program towards the end of the 60's. Description of MOLVIB: ----------------------------- Documentation for the program MOLVIB (version 6.0) Purpose: MOLVIB is a program for classical harmonic force field calculations on free and crystalline molecules. Language: Fortran 77 Hardware: VAX (VMS), IBM PC (DOS) Usage: Input instructions can be found in the supplied manual. A help file (in VAX/VMS help format) is provided. A standard test (ethane) is also included. Abstract: Normal coordinate analysis is nowadays commonly employed as an aid in the interpretation of the vibrational spectra of large molecules. In order to get meaningful results, a knowledge of the vibrational force field is necessary. Since the number of force constants grows quadratically with the number of atoms, one has to employ many approximations in the calculation of harmonic force fields even for moderately large molecules. To overcome this difficulty one can determine a force field for a set of related molecules using the so-called overlay method introduced by Snyder and Schachtschneider in the 1960's (J.H. Schachtschneider and R.G. Snyder, Spectrochim. Acta, 19 (1963) 117-168) About 1970 Gwinn developed a program for normal coordinate analysis using mass-weighted cartesian coordinates (W.D. Gwinn, J. Chem. Phys., 55, 477-481 (1971)), which eliminates the redundancy problems arising when internal valence coordinates are used, as in Wilson's GF-method. MOLVIB is based on the same fundamental idea, but differs from similar programs in many respects. The program was first described by T. Sundius, Commentat. Phys.-Math. 47, 1-66 (1977), and a more recent decription by the same author can be found in J. Mol. Struct. 218, 321-326 (1990). In addition to free molecules, crystals can also be treated. In this case, up to 50 atoms divided between 11 sub-units can be handled. All the calculations are performed using mass-weighted cartesian coordinates, instead of the conventional GF-method. This makes it possible to overcome problems with redundant coordinates. The force field is refined by a modified least squares fit of the observed normal frequencies, as described in T. Sundius, J. Mol. Spectrosc. 82, 138-151 (1980). Beginning with version 6 of the program, it is possible to change several of the maximum array dimensions, and create executables, which can handle larger molecules (or force fields). The program is user-friendly, and can be easily adopted to different force fields. It is also possible to express the force field in the CFF notation (see S.R. Niketic and Kj. Rasmussen: The Consistent Force Field: A Documentation, Lecture Notes in Chemistry, Vol. 3, Springer-Verlag (1977)). This makes it possible to use it in combination with the CFF program (or other molecular mechanics programs) for conformational analysis of flexible molecules. Accuracy: Single precision floating point. The matrix diagonalization can be performed in double precision, as shown by an alternative version of the routine EIGV. This may be advantageous on machines with small word length, especially if highly accurate eigenvectors are desired. The null frequencies, which always are printed out, can serve as a check both for machine accuracy and the validity of the input data. With single precision on a 32-bit machine, null frequencies of magnitude 1-2 cm-1 are not uncommon. Libraries: A few of the subroutines are modifications of programs previously published in books or journals, as also has been indicated in the prologues. If desired, these routines can be replaced by similar library routines (NAG or IMSL, e.g.), if such are available. Author: Tom Sundius, Department of Physics, University of Helsinki, January, 1991. ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==* From: case.,at,.scripps.edu (David Case) ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==* Not sure what you mean by a "normal mode computation procedure"....programs like Amber and Charmm and Discover all can compute normal modes of any molecule that their force fields cover (primarily but not exclusively biomolecules). I have a review of normal mode calculations on proteins that will appear in the April issue of Current Opinion in Structural Biology. ==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*==*