Go to the previous, next section.
There are number of parameter files available. Immutable versions are stored in the directory, `CGDATA'. The file names used for these files consists of an alphabetic part followed by a number, e.g. `PARAM5'. There are two copies of each file; one with extension, `.INP', which is a character files used as an command file to generate the binary file, with extension, `.MOD'. The `.INP' is meant for human eyes; the `.MOD' files is meant for CONGEN to read. The numeric part of each name is its version number. In general, one should use the highest version number of a file.
There are four types of parameter files available; protein, DNA, AMBER, and obsolete. The protein parameter files are called `PARAMn'. The DNA parameter files are called `PARMDNAn'. The AMBER parameters files begin with `AMBER', and the obsolete files are the other files which begin with `PARM'.
This appendix begins with an essay by David J. States describing the development of the `PARAMn'. The remainder of the appendix lists the titles from all the parameter files.
Parameters for proteins and Nucleic acids together
Obsolete Parameters for proteins
Parameters for DNA.
This essay describes the PARAM1 parameterization for the CHARMM molecular modelling program, compiled in May of 1981 by David States. My goal was to obtain a documented set of parameters which described the experimentally observable properties of small molecules as well as possible in the context of the CHARMM potential energy form. This was achieved by a combination of explicit least squares fitting where possible, and trial and error model building where it was not. Extended atom representations were used for aliphatic and aromatic hydrogens with the exception of benzene, where it was necessary to model the system using both extended and all hydrogen potentials. Explicit hydrogens were used for all protonated heteroatoms. Parameters have been included in the parameter file for atom types used in the PROT and ALLH topology files by generalizing the most appropriate HPRO parameter (ex. using the C-NH1 bond constants for C-NH1E).
The first step in the process of obtaining bond, angle, and dihedral parameters was to establish the mean bond and angle geometries. Bond lengths were taken from Bruce Gelin's Ph. D. thesis (the dipeptide modelling section), and agree quite well with other tabulations of standard bond lengths such as the CRC Handbook and Karplus and Porter. Angles were chosen as the average value observed in protein crystal structures available at better than 2.0 angstroms resolution on the Brookhaven Protein Data Bank tape. This includes the following structures:
These coordinates and their sequences were read off the tape and tables of the angle geometries built using the CHARMM program. The average values for each angle type were collected over all of these structures weighted equally. This compilation allowed the averaging of twenty to fifteen hundred occurrences for each angle type, with variances ranging from less than one degree to as much as fifteen degrees, and ranges of two to sixty five degrees. In general these average values are within two degrees of the `PARMFIX10' values. It should be noted that all of the above structures are crystallographically refined. A certain amount of bias is therefore inevitable, but I feel justified by the higher resolutions that these methods achieve.
Having established these geometric parameters, the next step was
the definition of force constants. Force constants describing simple
hydrocarbons and amides were obtained by least squares fitting to the
vibrational spectra of propane, butane, pentane, and N-methyl acetamide
(Snyder et al and Warshel, Levitt, and Lifson). This fitting was carried
out in stages, first choosing parameters and evaluating the resulting
normal modes until a set was obtained with correctly ordered vibration
frequencies. These parameters were then refined by least squares fitting
to vibrational spectra of all the isotopic variants available. This
fitting was performed with a user subroutine linked to CHARMM that
called the CHARMM energy and vibrational analysis routines for normal
mode analysis, and used an IMSL minimization routine (ZXSSQ).
This fitting algorithm uses finite differences to construct an
approximate first derivative vector and second derivative matrix,
requiring a rapidly increasing number of function evaluations as the
number of free variables rises. To reduce the complexity of this
fitting, parameters were grouped (bonds vs. angles vs. dihedrals and
impropers) and minimized separately until good agreement was obtained
over the entire range of observed frequencies. The experimental and
calculated normal modes for N-methyl acetamide are shown in the first
table. The most prominent differences between the current and the
previous parameterization are seen in the low frequency and out of plane
bending modes.
The dihedral angle potential for alkanes was found to be 1.6 kcal/mole by this procedure in good agreement with the value reported in the dipeptide and acetylcholine work of Gelin (1.8 kcal/mole). It should be noted that the previous parameter sets used a much lower value (0.2 kcal/mole). The old value resulted in essentially unhindered rotation between trans and gauche isomers and a very small gauche well depth (0.03 kcal/mole). The new parameters are in much better agreement with experimental observations as can be seen in the second table.
For the aromatic systems it was not possible to perform this least squares fitting because the simple potential energy form that we are using is not able to correctly order the various normal modes (Varsanyi and Dollish et al). This is apparently a result of the strong valence interactions between internal coordinates, and presumably could be overcome using a potential form that includes valence cross terms. The large number of modes (30) in benzene and the strong interaction of carbon and hydrogen motions also complicated the analysis of this system. For these reasons, benzene was analyzed by manually adjusting parameters for a model including all hydrogens. The heavy atom parameters obtained from this system agree with those used previously, and were kept fixed in modelling more elaborate aromatic systems.
Using the CHARMM potential energy function, aromatic out of plane bending behavior is determined by both the arrangement and parameterization of the improper dihedrals. For histidine, phenylalanine, and tyrosine the current parameterization and arrangement gave good agreement for the intrinsic ring modes (based on the benzene modelling), and they were retained. The ring substituent out of plane modes were not well described, and it was necessary to parameterize the system using the lowest out of plane bending modes for toluene (217 cm-1) and phenol (241 cm-1) to fix new constants for the ring CB and ring hydroxyl bends respectively. The present improper dihedral was retained, but in both cases the force constants required large increases (from 25 kcal/mol/rad2 to 120 and 150 respectively).
Tryptophan required both rearrangement and reparameterization of the improper dihedrals to obtain reasonable agreement with the skeletal out of plane bend (400 to 500 cm-1 in purines and napthalene) without creating very high frequency modes involving the ring intersection. The new arrangement replaces the improper dihedrals centered on the atoms at the intersection of the rings (CD2 and CE2) with terms that are trans two fold dihedrals beginning on one ring, running across the bridging atoms and ending on the opposite ring as shown below:
CZ2 NE1
* * * *
* * * *
CH2 CE2 *
| | *
| | CD1
| | *
| | *
CZ3 CD2 *
* * * *
* * * *
CE3 CG
|
|
CB
PARAM1 PARMFIXn
CZ2 CZ2----------NE1
* * *
* * *
CE2 * CE2
| *
| *
CD2 CD2
*
*
CG
NE1
*
*
CE2 CE2
| *
| *
CD2 * CD2
* * *
* * *
CE3 CE3-----------CG
In addition, two improper dihedrals were added as shown below.
CH2--------CE2 CE2 *
| | *
| * CD1 | * CD1
| * |
CD2 * CZ3-------CD2
This new arrangement distributes the out of plane bending forces evenly across the ring in a manner analogous to the delocalized electronic structure of the system, and does so without excessively elevating the frequency of CD2/CE2 out of plane modes. With the `PARMFIX10' parameters these modes had frequencies of about 1000 cm-1, but the skeletal out of plane bend occurred at 120 cm-1. Simply raising the force constants without rearranging the dihedrals resulted in a bridge atoms out of plane mode at 3000 cm-1! Aside from being unrealistic, this high frequency would force the use of small step sizes to retain accuracy in the dynamics integration.
My effort concentrated on the bond, angle, and dihedral force constants. The non-bond interaction parameters from Bruce Gelin's thesis were used without modification or cutoffs for the model systems. These nonbond parameters are the same as those found in `PARMFIX10'.
The hydrogen bond geometries and well depths were obtained from a table of typical values found in Vinogradov and Linell. Because the van der Waals and electrostatic terms contribute significantly to the total hydrogen bond energy, the hydrogen bond parameters were adjusted to give an energy minimized water dimer with the desired geometry and well depth. The other terms were shifted similarly, although by smaller amounts because the van der Waals interaction is not as large in the longer nitrogen-oxygen and nitrogen-nitrogen pairs. No hydrogen bonds were present in the model systems used above.
`PARAM1' is a refined set of parameters for the CHARMM potential energy form that describes the molecular mechanics of systems containing the covalent structures found in biological molecules. While an effort has been made to include different model compounds and conformations to generalize these parameters, they are primarily based on simple amides and hydrocarbons and intended for use in protein simulations. For non-protein systems it would be advisable to verify the properties of the constituents using model compounds (such as sugars and nucleotides) before macromolecular simulations are attempted. The accuracy of the geometric and force constants presented in this set is difficult to estimate. For some parameters such as bond stretching constants isolated modes could easily be identified and the parameter set to an accuracy of five percent or better. Conversely, angle bending and dihedral constants could not always be associated with particular modes, and significant correlations between the various parameters were observed during fitting. Even where least squares fitting was possible, none of these force constants should be regarded as accurate to better than twenty per cent. One corollary of this interdependence is the danger in naively readjusting parameters without repeating the analysis of all of the various model systems affected.
Given the number and extent of the changes presented in this work, the question of how they will affect results naturally arises. It should be noted that almost all of the changes presented here will tend to make molecules more rigid than they were previously. This will have implications in both the accuracy of the dynamics algorithm at a given step size, and the magnitude of the expected fluctuations. The behavior of the peptide bonds should not be markedly changed since the net effect of the parameter change on their vibrational modes is small. The aliphatic sidechains and coupling of motions through angle and dihedral terms is expected to show more prominent effects.
Dollish, Fately, and Bentley (1975) Characteristic Raman Frequencies of Organic Compounds Wiley and Sons, New York.
Gelin Ph. D. Thesis (Chapters 2 and 4, Tables 12 to 15)
Gelin and Karplus (1975) JACS 97: 6996
Snyder and Schachtscheider (1965) Spectrochimica Acta 21: 169-195
Varsanyi (1974) Assignments to Vibrational Spectra of 700 Benzene Derivatives Halsted Press, Budapest.
Vinogradov and Linell (1971) Hydrogen Bonding van Nostrand.
Warshel and Lifson (1970) J. Chem. Phys. 53: 582
Warshel, Levitt, and Lifson (1971) J. Mol. Spect. 33: 84
mode Experimental New Old
____________________________________________
peptide torsion 192 189 162
N bend 289 281 291
C in plane 437 444 375
C out of plane 600 551 312
Amide IV 628 621 544
NH out of plane 725 730 390
C C stretch 883 866 864
C N stretch 1120 1115 1112
Amide III 1323 1314 1262
Amide II 1570 1597 1580
Amide I 1660 1661 1675
N H stretch 3200 3193 2969
Angle Experimental New Old ____________________________________________________________ 180.0 0.0 -0.23 -0.23 165.0 0.21 0.00 150.0 1.32 0.27 135.0 2.49 0.65 120.0 3.4 3.04 0.86 105.0 2.69 0.87 90.0 1.82 0.84 75.0 0.8 1.32 1.05 60.0 1.67 1.65 45.0 2.97 2.61 30.0 4.92 3.68 15.0 6.73 4.54 0.0 6.1 7.43 4.87
Term New Set Old Set
____________________________________
C O bond 580 600
C CH bond 405 400
C NH bond 471 500
NH CH bond 422 450
NH H bond 405 350
CH C O angle 85 40
CH C NH angle 20 25
NH C O angle 65 50
C NH CH angle 77.5 60
C NH H angle 35 30
CH NH H angle 30 30
CH CH CH angle 45 30
C NH torsion 8.0 7.0
CH CH torsion 1.6 0.2
C improper 105 25
N improper 50 25
toluene improper 120 25
phenol improper 150 25
The values for CM same as that for carbonyl carbon, and the values for OM same as that for carbonyl oxygen.
The AMBER parameter set(38) is available as `AMBERPARM'.
This parameter file is now obsolete, and is replaced by the AMBER 94 parameter file, which is described in the next section.
The AMBER 94 parameter set(39) is available as `AMBER94PARM'. This parameter file must be used with the AMBER 94 topology file, see section AMBER94RTF.
Several atoms in the AMBER94 potential have zero van der Waals radii. This causes problems for CONGEN because of the possibility that such atoms may get close enough to another atom for electrostatic fusion to occur. Therefore, such atoms have been given a small van der Waals radius and a small van der Waals repulsive interaction. Please see the file for the exact value.
This potential file has some features which you should be aware of. First, the amide torsion terms (as found in the aspargine and glutamine sidechains) are designed so that the minimum energy is non-zero, and the shape is correct only when both hydrogens are considered. Second, there is no explicit hydrogen bond energy. Such interactions are accounted for by the electrostatic energy. Third, zero energy improper torsion terms have been added for chiral centers so that the analysis tables, see section Static Properties of Internal Coordinates, will show the chirality and allow the user to check it.
Title found in the run creating the parameters:
* PARAMETER FILE FOR EXEL VERSION 11 CREATION RUN * HYDROGEN BOND STRENGTHS INCREASED BY 1.0 KCAL/MOLE FROM PARMFIX11 * LENGTHS DECREASED BY 0.1-0.2 ANGSTROM WALLWORK ACTA. CRYST. (66) 15:758 * 7-Apr-1981 21:33:52 - DJS
Title written to the binary file:
* PARAMETER VERSION 10 -- EXTENDED AND EXPLICIT SET -- 27 ATOM TYPES * COMPILED BY WFVG, SNS AND BDO -- MARCH 1979 * HYDROGEN BOND PARAMETERS CHANGED FROM VERSION 5 -- ONLY O-O,O-N,N-N * H BOND STRENGTHS INCREASED 1.0 KCAL/MOLE FOR EXEL LENGTHS SHORTENED * NONBONDED PARAMETERS FOR HYDROGEN (A-0.044,N-1,R-0.8 == EMIN-0.1) * PHIS INVOLVING C(CARBONYL) AND NITROGENS MADE CONSISTENT IN VER. 8 * O-H BOND FORCE CONSTANTS CHANGED TO 400.0 IN VERSION 9 * ADDED ANGLE FOR N-TERMINAL PROLINE -- BRUC * LAST UPDATE 7-Apr-1981 21:20:20
Title written to the binary file:
* PARAMETER VERSION 10 -- EXTENDED AND EXPLICIT SET -- 27 ATOM TYPES * COMPILED BY WFVG, SNS AND BDO -- MARCH 1979 * HYDROGEN BOND PARAMETERS CHANGED FROM VERSION 5 -- ONLY O-O,O-N,N-N * NONBONDED PARAMETERS FOR HYDROGEN (A-0.044,N-1,R-0.8 == EMIN-0.1) * PHIS INVOLVING C(CARBONYL) AND NITROGENS MADE CONSISTENT IN VER. 8 * O-H BOND FORCE CONSTANTS CHANGED TO 400.0 IN VERSION 9 * ADDED ANGLE FOR N-TERMINAL PROLINE -- BRUC * LAST UPDATE 16-Mar-1981 21:54:36
Additional angle for proline cropped up when a bug was found in
PATEXH and explicit hydrogen topology file. The angle added was
CH1E-NH3-CH2E which was given the parameters for HC-NH3-CH1E which
appeared to be the closest angle to it. --Bruc
Title written to the binary file:
* PARAMETER VERSION 9 * EXTENDED AND EXPLICIT SET -- 27 ATOM TYPES * COMPILED BY WFVG, SNS AND BDO -- MARCH 1979 * HYDROGEN BOND PARAMETERS CHANGED FROM VERSION 5 -- ONLY O-O,O-N,N-N * NONBONDED PARAMETERS FOR HYDROGEN (A-0.044,N-1,R-0.8 == EMIN-0.1) * PHIS INVOLVING C(CARBONYL) AND NITROGENS MADE CONSISTENT IN VER. 8 * O-H BOND FORCE CONSTANTS CHANGED TO 400.0 IN VERSION 9 * LAST UPDATE 5/4/80
Title written to the binary file:
* PARAMETER VERSION 8 * EXTENDED AND EXPLICIT SET -- 27 ATOM TYPES * COMPILED BY WFVG, SNS AND BDO -- MARCH 1979 * HYDROGEN BOND PARAMETERS CHANGED FROM VERSION 5 -- ONLY O-O,O-N,N-N * NONBONDED PARAMETERS FOR HYDROGEN (A-0.044,N-1,R-0.8 == EMIN-0.1) * PHIS INVOLVING C(CARBONYL) AND NITROGENS MADE CONSISTENT IN VER. 8 * LAST UPDATE 3/23/80
Title written to the current binary file:
* PARAMETER VERSION 7 * EXTENDED AND EXPLICIT SET -- 27 ATOM TYPES * COMPILED BY WFVG, SNS AND BDO -- MARCH 1979 * HYDROGEN BOND PARAMETERS CHANGED FROM VERSION 5 -- ONLY O-O,O-N,N-N * NONBONDED PARAMETERS FOR HYDROGEN (A-0.044,N-1,R-0.8 == EMIN-0.1) * LAST UPDATE /1/24/80
Title written to the binary file:
* PARAMETER LIST - VERSION 5 * EXTENDED AND EXPLICIT SET -- 27 ATOM TYPES * COMPILED BY WFVG, SNS AND BDO -- MARCH 1979 * LAST UPDATE 4/10/79
Title written to the binary file:
* EXTENDED ATOM PARAMETERS -- VERSION 2 * 15 ATOM TYPES -- THIS PARAMETER FILE WAS THE MOST LIKELY * FILE WITH WHICH THE LONG ( 100 PICOSECS ) PTI DYNAMICS RUN * WAS MADE -- IT HAS SOME LARGE ANGLE FORCE CONSTANTS AND NO * PARAMETERS FOR IMPROPER TORSIONS -- PARAMETERS NOT NEEDED IN * PTI HAVE ALSO BEEN ADDED -- 3/15/79 -- BDO AND SNS
This file is obsolete. The latest version of `PARAMn' should be used for all hydrogen topology files.
Title in the input file generating the binary file.:
* PARAMETER FILE FOR ALL EXPLICIT HYDROGENS BASED ON CFF PARAMETERS * IN BRUCE GELIN'S THESIS (PARMFIX7 VALUES HAVE TRAILING 0'S) * BOND ANGLE PARAMETERS INCLUDE THE HARMONIC CONTRIBUTION FROM UREY-BRADLEY * F TERM ( (R13-R0)**2 ) * DIHEDRALS INCLUDE MULTIPLICITY FROM CFF * MODIFIED TO INCLUDE HA AND CT * HIS AND TRYP INCLUDED 4/13/80 DJS
Title written to the parameter file:
* ALLH PARAMETER SET MODIFIED FROM GELIN THESIS 3/4/80 * MODIFIED TO INCLUDE HA AND CT -- 2/24/80 -- DJS
The DNA parameter files have had very limited usage. Do not trust the results using them unless you satisfy yourself that the parameters are reasonable.
The DNA parameters are inconsistent with the protein parameters. Do not mix proteins and DNA together. You will run into problems with chemical type codes being reused and general confusion and errors.
`PARMDNA4' is the same as `PARMDNA3' except a sodium ion has been added to the nonbonded parameters.
Creation title:
* PARAMETER FILE -- VERSION 4 -- CREATION RUN -- 10/20/81 -- BRB
Title in the file:
* PARAMETER VERSION 4 * DNA
Titles from the creation run:
* PARAMETER FILE -- VERSION 2 -- CREATION RUN -- 10/20/81 -- BRB * BROMINATED CYT AND WATER ADDED 10/20/81
Title in the file:
* PARAMETER VERSION 2 * PARAMETERS FROM UCSF EXCEPT FOR HBONDS AND IMPHIS * DNA
Title from the input file:
* PARAMETER FILE -- VERSION 1 -- CREATION RUN -- 1/29/81 -- BRB
Title written to the binary file:
* PARAMETER VERSION 1 * PARAMETERS FROM UCSF EXCEPT FOR HBONDS AND IMPHIS * DNA
Go to the previous, next section.