Validation of the General Purpose QUANTAa3.2/CHARMma Force Field" Frank A. Momanyt and Rebecca Rone Polygen/Molecular Simulations, 200 Fifth Avenue, Walth,am, Massachusetts 02254 Received 9 September 1991; accepted 24 March 1992 An evaluation of the CHARMm force field for small molecules is described. Using different force field conditions and computational techniques, a wide variety of compounds are analyzed. rms deviations of Cartesian coordinates for 49 diverse organic molecules taken from the Cambridge Crystallographic Data Base and internal coordinate geometries for 28 other molecules are reported. Results are described with different dielectrics, dihedral constraints, and crystal packing to allow analysis of deviations from experimental data and give precise statements of the reliability of the parameters used in the force field. Torsional barriers (rms = 0.4) and conformational energy differences (rms = 0.4) are examined and comparisons made to other force fields such as MM2, Tripos, and DREIDING. The results confirm that CHARMm is an internally consistent all purpose force field with energy terms for bonds, angles, dihedrals, and out-ofplane motions, as well as nonbonded electrostatic and van der Waals interactions. Reported CHARMm results (rms = 0.006 A for bonds, rms = 1.37" for angles, and rms = 3.2" for dihedrals) are in excellent agreement with high quality electron diffraction data. 0 1992 by John Wiley & Sons, Inc. INTRODUCTION Molecular mechanics calculations are widely used for the study of small organic molecules, proteins, sugars, lipids, and other biopolymers. More recently, calculation methods have become available for large polymeric systems, many inorganics, liquids, ionic solutions, and crystal lattices. The CHARMm force field' consists of a functional form for the potential energy th at lends itself relatively easily to the study of a variety of different types of molecules. Individual energy contributions are obtained from functions that are simple and can be interpreted in a physical sense. These include harmonic bonding and angle terms, cosine torsional or dihedral terms, and nonbonded terms which include electrostatic Coulombic terms using partial atomic charges and van der Waals Lennard-Jones 6-12 contributions. In this article, CHARMm results are compared to experimental data and to reported results from other force field^.^,^ In the evaluation of th e results several features of the use of selected force fields will be noted. For example, th e inclusion of full electrostatic Coulombic interactions using partial atomic charges is standard in CHARMm, and *QUANTAAND CHARMm are registered trademarks of Polygen Corporation. QUANTAKHARMm parameters are from Release 3.2, Polygen Corporation. ?Author to whom all correspondence should be addressed. was used in this study. However, in one study' used for comparisons in the discussion section, the authors chose not to include electrostatics in their reported results. When the electrostatic terms are not included in the force field, one must be particularly cautious in interpreting simulations on polar molecules and their intra- and intermolecular interactions. Improving the parameters that are used within the CHARMm force field is a n ongoing project and hundreds of molecules have been analyzed in the development of current parameter^.^ The molecules used in parameter development include saturated hydrocarbons, unsaturated hydrocarbons, strained systems, conjugated and aromatic systems, all major polar functional groups, including alcohols, aldehydes, amines, etc., as well as heterocyclic and ionic systems. The result is a force field that is applicable to small organic molecules, or large polymeric molecules, as well as macromolecular biologically related molecules. The experimental data used in parameterization includes gas phase electron diffraction and solid state molecular structures as well as high level SCF ab initio calculations. The ab initio results are very useful in giving conformationally dependent geometry trends that are not readily available from experimental measurement^,^ energy differences between conformations in selected comparisons,6 and barriers to rotation about dihedral angles6 Where sufficient ab initio basis sets are included Journal of Computational Chemistry, Vol. 13, No. 7, 888-900 (1992) 0 1992 by John Wiley & Sons, Inc. CCC 0192-8651/92/070888-13$04.00 This and many other cases are parameterized to result in subtle but important structural enhancements. Dihedral angles were constrained at the X-ray values and internal coordinate geometry was energy minimized to validate geometries without the complications of conformational change. In the analysis of crystal structures from the Cambridge Crystallographic Data Base. includes full energy mini1 mization of each conformation. This list includes 102 organic atom types and many metals in the periodic table that have had their nonbonded terms parameterized. or to the appropriate charge of an ionic system. the data has been used to refine van der Waals and electrostatic nonbonded interaction^. DREIDING. Different atom types for the same element were historically introduced to reflect variations in orbital hybridization arising from resonance or conjugated states. The 2D structure is converted to 3D. Different dielectric values were used to test the effect of the electrostatic atom-atom interactions on conformation. One series of molecules used in the comparisons was taken from the Cambridge Crystallographic Data Base. and assign partial atomic charges to each atom. out-of-plane. etc. the molecules were simulated in their experimental crystalline environment as specified by their space group to examine the effect of the crystal forces on the conformational energetic preference. as shown in Tables I and 1 . Electrostatic energies on atoms separated by three bonds (1-4 interactions) are scaled down by a factor of 1/2. The torsional energy barrier study reports the CHARMm. Because of this effect. such as in the case of alcohols where the rotation about the hydroxyl C-0 bond causes variations in the 0-C-H angle of the methylene group. Other special 1-4 terms have also been found to be important. The parameter file for CHARMm contains 6-12 nonbonded parameter terms for all the listed atom types. UR). and again atom types and partial atomic charges are assigned to every atom. and nonbonded interactions. Large rms deviations in atom positions can result from small variations in one or more dihedral angles when several descendent groups are moved by this small change.^ Parameter refinement and the determination of the charge set’ will not be described here since this article is directed to the preliminary validation of results of calculations using the force field. That reasoning has continued today with new atom types included in QUANTA 3. No special hydrogen bond terms are included.g care is taken in this article to fully understand the cause of calculated rms deviations from experiment. Special parameters are included for specific 1-4 interactions across a bond. The nonbonded van der Waals 1-4 terms are not scaled but some specific atom types are modified as determined to be necessary by conformational energy studies.2. angle bend. dihedral. electrostatic 1-5 or greater. including bond stretch. assign atom types. Other molecules studied were drawn in CHEMNOTE (a 2D drawing routine within QUANTA 3. Atom typing is carried out using: (1)bond types for bonds connected to the atom. This scaling was found to be necessary to minimize the difference between conformational energies when using different dielectric methods (C = 1 vs. the molecules examined here were studied using a variety of conditions and techniques.2 to enhance the reliability of new and uniquely different functional groups. MP3.’ However.2. In selected cases. Charges are derived from a charge template where every atom has a charge assigned to it depending upon the microenvironment of atom types as found from the bonding path algorithm. The column designated “rigid” under CHARMm is obtained by simply rotating the dihe- .g After crystal data files are read the QUANTA molecular editor is used to place hydrogens. The van der Waals terms are provided for homonuclear atom pairs and standard combination rules are applied for the energy (geometric mean) and distance (arithmetic mean) of the heteronuclear pairs.2). such as in the cases of extended atoms. electrostatic 1-4. (2) element types of the central atom and elements connected to it.) terms calculated. METHODS The CHARMm force field equations are identical to those reported in the original CHARMM publication. The CHARMm energy differences include all contributions to the energy. Charge smoothing algorithms are used to net the total charge to zero for neutral molecules. The method for defining the dihedral barriers to include torsional terms for every set of four atoms across the bond has recently been adopted in QUANTA 3. the parameters used throughout this study are those distributed by Polygen in QUANTA Release 3.3 and Tripos’ results obtained by holding the dihedral angle constrained and relaxing the geometry by energy minimization (see the relaxed column under CHARMm). The method used to study the torsional barrier heights and conformational energy differences. The partial atomic charges are designed to give correct interaction energies between molecules and groups within molecules and to give correct interaction distances such as found in hydrogen bond distances between polar molecules.CHARMM FORCE FIELD 889 and correlation (MP2. and (3) a bonding path algorithm for searching its microenvironment. and are available upon request from Polygen/Molecular Simulations. o 0.5 "All comparisons are made with the experimental values in column 1.9' 3.8 0.5 0.3 3.4 1.7 3.7d 1.4 0.5 0.2 0. Comparisons of experimental torsional barrier heights (kcalimol) with those obtained from Tripos.8 DRIEDING (3) 0.o 3. Table 11.3 1.6 1.5 -b 0.7' 3.3 2.3 2.9 3.8 -h -b CHARMm 0.1' 0.2 1.1 3.0d 1.6 0.3 3.7 5.4 0.9 2.890 MOMANY A N D RONE Table I.9 4.4 - 3.4 3.9 3.5 4. .7 0.9 2.6d 3.9 3.6 1.9 0.8 3.3 2. The rms when this value is included is 0.9 4.6 2. l C 2.5.180) 1.1 3.4 2.3 3.3 0.4 1.1 4.7 0.6 5. dReference 12.5 3.1 4.2 0. and CHARMm force fields.9 3.8 4.1 6.8 4.6d 1 .0 1.6 0.4 0. Comparisons of experimental energy differences between conformers with those obtained from the MM2.6 0. 1-Trichloroethane Hexafluoroethane Dimethylthiol Propene (CH3-CHCH2) Acetaldehyde Acetic acid (CH3-COOH) rms deviations 3.3 4.0 MM2 (2) 0.0d 3.0 4.4b 1.5 1.4 3.5 1." CHARMm TRIPOS BIOGRAF Moleculeb Experimental (2) (3) Rigid Relaxed Set 1 Ethane Propane Butane.2 0. twist-boat-chair rms deviation Experimental (11) 0.7 3. a-e Methyl-ethyl-ether. 'Reference 11.8 4.3 1.4 3.2 2.5 1.1 4.8 2. a-e Chloro-cyclohexane.3 3.9 5.0 0.4 0.7 -0.0 3.0 5.4-Dichlorocyclohexane.4d 6.0 1.5' 2.9 "Ab initio value in parentheses is from reference 10. a-e Nitrocyclohexane.4 0.7 0.7' 3. g-t 2-Methylbutane.1 7.0d 3.7 1.8 0.9d 4.6 1.8 1.6 0.9d 2.3 0.2' 5.9 0.2 2.1 2.5 6.9 3. a-e Phenyl-cyclohexane.8 0.5)a 2. Tripos. Nitro-cyclohexane was omitted in the CHARMm calculation since data was not available for comparison in the MM2 results.9 1. Molecule Butane.3' 3. l .7' 3. g-t Propionaldehyde (120-0) Cyclohexane.1 1.7 1.5 79.2 (-0.0 0.4 1.6 0.0 5.1 3.2 0.2 0.1 4.7d 2.8 2.1.6 0.2 1. a-e Fluoro-cyclohexane.4 0.8 4.0 4.5 3.1' 3. a-e(CCOH.9 0.9 5.3 0.8 1.2 1.6 2. a-e Bromo-cyclohexane. Me-Me eclipsed Butane.2 3.3 1.0 0.1 3.4 TRIPOS (2) 0.8 -0.8 4.0 0.6 -b 7.4 4.9 3.2 0.5 0. bSet 1 includes molecules in the Tripos article.0 -0.1-Trifluoroethane l .1' 2.9 0. g-t 1-Butene (120-0) deg Butadiene (40-180) deg Methyl-cyclohexane.4d 3.7 1.4 -h 1.0 3.3 1. bOmitted from calculation of rms.1 2.4h 1.3 0.1 1.2 while set 2 includes more examples found in the BIOGRAF3 article.7 2.9 6.4 0. BIOGRAF.5 3.7 5.9 1.4 0.0 5.o 1.0 0. a-e Cyclohexanol. and CHARMm force fields. Me-H eclipsed 2-Methyl-butane Neopentane Methylsilane Methylamine Dimethylamine Fluoroethane Chloroethane Bromoethane Iodoethane Methanol Dimethylether rms deviations Set 2 1.3 2.5 4.1 3. Where possible the comparisons to CHARMm were made with the newer MM3 results of Allinger. Thus.1° Results using CHARMm and the MM2 results are in agreement with this a b initio data. while the angles are of the order of a degree (rms of 1.7-6. For example. in the case of cis-butane where two methyl groups are cis across the bond.CHARMM FORCE FIELD 89 1 dral angle from the lowest energy minimized structure and calculating the energy at particular values of the dihedral angle close to the energy maximum. and the release of strain energy is in these cases important.7 kcal/mol) and the relaxation of geometry and release of strain takes place primarily by increasing the bond angles on each side of the bond of interest. but was chosen to allow comparison with previously published calculation^. These results of Table I1 are also of interest for comparison purposes. and dihedrals) obtained from energy minimization of the molecules of Tables I and T are shown in Table 111.3and CHARMm in Table 11.was obtained for all single molecule studies and 1 x for the crystal studies. For example. In some cases MM2 comparisons were necessary since the newer MM3 results were not yet available in the literature.^. are found to be calculated to within several degrees using CHARMm (rms of .2 DREIDING. The saturated and unsaturated hydrocarbons are well represented and give excellent agreement with experimental studies and compare well with MM3 results. Experimental dihedral barriers are compared with calculated results of Tripos. CFF32 calculations. one finds real chemically significant reasons for the difference. The energy difference between relaxed and rigid geometries in this case is significant (4.^^^ The newly implemented multiple periodicities of torsional terms in CHARMm results in very reasonable barriers for both geometry relaxed and rigid structures. plus multiple periodicities for the torsional barrier terms. the CHARMm calculated bond lengths are found to be within several thousands of an A (rms of 0. This limited list of compounds was chosen for comparison purposes with other published force field^. I These coordinates are compared to results from MM234. which are the hardest internal coordinates to obtain correctly from calculations. In the single molecule calculations cutoff distances were taken to be greater than the longest intramolecular distance. considering those structures where comparisons with MM3 and experiment can be made. 0. Dihedral angles. where the lowest energy conformation was found to occur near a dihedral angle of 120" according to recent ab initio results. where the atoms 1-4 across the bond contribute to the torsional barrier significantly. A RESULTS AND DISCUSSION Conformational Energy Differences and Torsional Barrier Heights The Tripos (and their reported MM2 results). scaled 1-4 nonbonded and electrostatic interactions. and that it is a balance of forces that contributes to these torsional barriers. one finds steric hindrance between methyl groups. A gradient over all atoms of better than 1 x l o p 6 kcal/mol. experimental values were used in computing rms differences. Geometries of these molecules were not available in the Tripos2 or DREIDING3 articles. MM313. In the few cases where the differences between the relaxed and rigid values are significant. BIOGRAF has a higher rms than the CHARMm results. the BIOGRAF DREIDING3 dihedral barriers are all set to 2.0 kcal/ mol for single bonds. symmetry copies within 15 A of any atom in the reference molecule were generated and used. This suggests that there are no major strain forces act- ing across 1-4 interactions.005 A for MM3).The dielectric constant was maintained at C = 1 for all calculations (Tables 1-111) with the exceptions noted in Table IV. Energy minimization was carried out using the Newton-Raphson method for single molecules and conjugate gradient method for crystal studies.' DREIDING.^ The comparisons show clearly that MM2 and CHARMm results are comparable. It is because CHARMm includes the microenvironment. An anomaly is found for the case of 1-butene.3and CHARMm conformational energy differences are compared to the experimental energy differences for gauche-trans and axialequitorial conformations in Table I. angles. As shown in the second set of more complex molecules. Crystal calculations used interaction cutoffs of 15 A with a switching function' between 11-14 A. the rms of MM2 and CHARMm will be improved. Nevertheless. From the space group information. If the 1-butene experimental data is in error. Crystal packing studies on selected molecules were carried out using space groups reported in the Cambridge Crystallographic Data Baseg for the particular molecule. Internal Coordinate Geometries Internal coordinates (bond lengths.0. Lattice parameters were held constant and the reference molecule and its symmetry copies allowed to change during energy minimization.37". For example. their results are very good.006 A. and electron diffraction or microwave data (see Table 111). that the overall agreement with experiment is better than force fields missing these terms for the different types of complex molecules studied. The data set is limited. in the first set of data shown.81 for MM3). 9 MM3I3 1.4 113.461) (123.6 ED^^ MM320 1. 111. and dihedral angles of molecules of Tables I and 11.534 111.468 122.544 2-Methylbutane c-c c-c-c 1.4 111. 111.536 1.539.532.531 1.8 Cyclohexane c-c c-c-c c-c-c-c Methyl-cyclohexane(e) c-c C(Me)-C c-c-c c-c-c-c 3-4 #1-2 2-3 1.467 124.534 1.9 Unsaturated hydrocarbons ED^^ 1.4 113.0. 1.6) * z tl Butadiene z a c=c c-c c=c-c .533.535 112.538 1.344 1. 112. 109. MM313 1. 1.534 111.5 ED14 1.0 Propane c-c c-c-c ED'^ 1. Calculated CHARMm.5 111.3 67.53l(ave) 1.534.534.5 4-3 5-3 1. 1.0 n-Butane C--(anti) c -C(g) C -C.Table 111.538 112. 1.5.4 1.536 1. 1.535.530 1.541 EDI~ 1.9 ED" 1.4 (1.535 234(t).6.344 1.537 EDI7 1. 234(g).4 MM313 1.533.536 111.1 #1-2 2-3 1.536 111.C( anti) c -c-C(g) c -c -c -c ED'^ 1. CCF.536 110. or ab initio geometries and experimental gas phase electron diffraction (ED) bond lengths.C -H(c) MMP 1.3. 1. 113.2 55.0 111.9) ED^^ (1.536.0 1.1 54. bond angles.534 #123 234 345 111.8. 112.4 1.2.4 Neopentane c-c MM313 1.532.538 111.1.533 111. 110. 435 110.4(ave) 71 Ethane C-C(t) c-C(c) c-c-H(t) C.534 112.4 54.349 1.531 112. or CFF Experimental Molecule Bond angle dihedral Saturated hydrocarbons #1-2 2-3 1. CHARMm MM2. 1. MM2.537 #4-5 1.535.340) (1. MM3.4.344) (1.467) (122.540 1.53l(ave) 113. 111. MM3.533 112.3 MM3I3 1. 1.461 122.4 55.4 55.7 64.4(ave) 113. 1. #123. 412t.0g 112.422 W c-0-c 0-c-c c-c-0-c 1.970 C-C-0-H ED^* 1.423 0.3 119.413 1.4t.5339 1.8 Dimethylether 1.0 c-0 c-0-c Me-tilt Ethyl-methyl-ether c-c C(Me)-0 C(CH2)-0 ED31 1.6 125.428 0.9 109.535 Phenyl-cyclohexane c-c b C. 111.531 123.506 1.421 3.69 79.9 c=c-c c-c-c c =c -c -c MM324 1.7 74.412 114.2 80.410. 1.9 120.530 1.0 w .502 1. 1.414g 1.3t.7 108.531t. 1. 115.6 178.5 EDz5 1.431 0.C(Ar) C -C -C(Ar) C-C(Ar)-C(Ar) c-c-c c-c-c-c C-C-C-C(Ar) C -C -C(Ar).345 1.415 111.428 0.535.5 CFF~~ 1.1 108.405 111.176. 1.0 MM327 1.3.6 MM327 1.9 1. 1.412.948 108.6 23 34 1.534 124.4 1.8.421 0.945 108.438 0.339 1.9 62.4 110.7 119.3 Methanol c-0 MWZ9 1.3 (ave) 111.517 1.1 1.4 55. .5 Alcohols #12 1.6 111.520 1.418 1.4 84.3 112.415g 114.4 111.5 0-H C-0-H Ethers ED3" MM3z7 1.3 108.6 111.5 111.1-Butene C=C( 1-2) C -C(2-3) C-C(3-4) 1.960 109.C(Ar) C(Ar).538 Cyclohexanol(e) c-c c-0 0-H c-c-c c-c-0 C-0-H c-c-c-c c-c-c-0 55.340 1.9 114.526 1.964 107.964 111.418 1. 110.9 107.542 1. 1.C(Ar) #12 23 34 1.505 1.lg 109.534.4 56. 387 109.c 1 1.Table 111.207 4-21G36 1.4 1.786 109.520 1.210 124.532 1.5 111.518 1.531 2.1.4-Dichlorocyclohexane c-c c-c1 c-c-Cl c1.169 110.950 111.7 177.399 Chloroethane c-c c-c1 c-c-c1 1.340 112.5 124. MM3.l-Trichloroethane 1.503 1.151 MW34 1.9 99.194 124.3 ~ ~ 2 3 4 Halogens 1.512 1.3 C-F C-C-F F-C-F 1.816 98.805 Experimental Dimethyl-thiol c-s c-s-c 1.532 1. or CFF CFF32 1.371 109.8 ~ ~ 2 3 4 ~ ~ 2 3 4 1.534 1.788 110. 1.4 MW34 1.5 ~ ~ 2 3 4 Bond angle dihedral MM2.0 1.531 1.821 98.2 Iodoethane c-c c-I c-c-I 1.530 1.4 Aldehyde ED37 1.786 111.787 111.510 1.1 EDZ3 1.c.4 1.535.538 1.1 1.788 Fluoroethane 109.371 109.1.9 109.532 2.c.538 1.2 Acetaldehyde c=o C(Me)-C C(Me).497 1.3 1.7 111.533 1.0 Bromoethane c-c C-Br C-C-Br 1.c 1 c-c c-c1 c-c-c1 c-c-c-c c.c.949 111.530 1.5 c-c C-F C-C-F ED% 1.LTrifluoroethane c-c ED35 1.929 110. (continued) CHARMm 00 co l b Molecule 1.8 1.512 1.C =0 .1 55.6 109. 014 112.201 C(Me)-C 4-21@6 1.514 1.223 111.4 CFF39 1.214 MW4' (1.4 Methylamine C-N H-N C-N-H H-N-H MM342 1.5 c-c c=o C( Me). 1.020 112.531 1.328 1.896 1. 110.337 1.4 Nitrocyclohexane 1.2(skew) 124.4 108.9 122.106 MW43 1.514 1.9 ED44 1.6 126.8 117.357) (1.7 Methylsilane C-Si Si-H C-Si-H Nitrate ED46(CHz)zCHN02 1.537 125.569(skew) 1.209 113.465 Silane 1.1 125.463 1.1 105.194 111.210 1.ED^^ Propanal(skew) 1.209) c-c c-0 c=o H-0 c-c-0 c-c=o (1 12.5 1.960 108.9) 112.0) (126.520 1.484 110.532 1.2 124.506 1.474 1.515 1.6 3-21G45 1.6 -61.1 N=O C-N c-c O=N=O O=N-C C-C-N C-C-N=O .514 1.9 Acid Acetic acid ED4" 1.0 107.3 106.517 125.364 1.0 112.453 1.C -C c-c=o 1.9 110.877 1.2) H-0-C Amine 1.486 110.521(syn).8(syn).6 (105.515 1. 1.535.5 109.226 1.4 125.494) (1.190 0. 067 0.035 0.091 0.019 0.106 0.266 0.029 0.262 0.024 0.750" 0.492 0. b.028 0.084 0. All heavy atoms (rms) Compound AAXTHP ABAXES ABINOROB ABTOET ABBUMOlO ABZTCX ACADOS ACAFLR ACANILOl ACARAP ACBNZAOl ACBUOL ACCITRlO ACDXUR ACENAP03 ACFPCH ACFUCN ACGLSP ACGLUAll ACHGAL ACHIST20 ACHNAPlO ACHTARlO ACIMDC ACINDN ACINST ACKYNU ACMBPN ACMEBZ ACMTDE ACNORT ACNPAClO ACNPEC ACPENC i0b ACPPCA ACPRET03 ACPYNS ACSALAOl ACSESOl ACTAND ACTHBZ ACTHCP ACTOLD ACTYSN ACURID ACVCHO ACXMOL ACXMPR ADENOSlO Ave.023 0.379 0.046 0.034 0.422 0. rms movements ( A ) for: (a-f) all heavy atoms.215 0.448 0.564 0.211 0.072 0.051 0.268 0.338)' DREIDING (c) 0.341 0.040 0.025 0.048 0.099 0.071 0.170 0.089 0.511 0.025 0.046 0.139 0.152 0.047 - 0.263 0.216 0.568 0.668 0.192 0.057 0.151 0.175 0.184 0. and (9.334 0.178 1.294 0.133 0.231 0.310 0.123 0.072 0.126 0.049 0.017 0.553 0.056 0.070 0.010 0. CHARMm (a) 0.450 0.309 0.466 0.264 0.030 0.449 0.069 0.070 0.297 0.035 0.045 0. c) single molecules.073 0.224 0.412 0.821 0.252 (0.280 0.082 0.036 0.448 0.448 0. h) rings only.030 0. .466 0.548 0.040 0.077 0.045 0.209 0.355 0.073 0.349 1.108 0.275 0.028 0.896 MOMANY AND RONE Table IV.020 0.117 0.027 0.399 0.219 0.274 0.246 0.295 0.229 0.268 0.317 0.121 0.041 0.113 0.263 0.201 0.028 0.031 0.092 0.611 0.025 0.088 0.018 0.114 0.341 0.109 0.203 0.020 0.178 0. ~ arms is average of dimer set included in unit cell.170 0.146 0.032 0.124 0.463 0.259 0.202 0.122 0.800 0.093 0.312 0.090 0.047 0.054 0.118 0.254 0.371 0.050 0.020 0.066 0.401 0.671 0.305 0.595 0.262 0.247 0.075 0.109 0.033 0.236 0.240 1.086 0.078 0.299 0.408 0.047 0.033 0.493 0.568 CRYSTAL (el 0.409 CONSR (f) 0.125 0.225 0.175 0.414 0.293 0.097 0.432 0.257 0.802Y TRIPOS (b) 0.026 0.050 0.208 0.029 0.340 0.334 0.155 0.055 0.125 0.354 0.351 0. 'rms includes only those molecules in column (e).036 0.122 0. (d) 4R dielectric.211 0.020 - 0.022 0.087 0.051 0.074 0.459" 0.046 0.088 0.049 0.379 0.146 0.067 0.694 0.500 0.095 0.288 0.844 0.391 0.061 1.312 0. (e) crystal (CRYSTAL) environment.290 0.025 0.795 0.148 0.233 f0.112 0.096 0.121 0. (a.175 0.025 0.012 0.140 0.091 - - 0.420 0.117 0.126 0.015 0.180 0.040 0.049 0.227 0.257 1.040 0.265 0.220 0.147 0.297 0.020 0. dev.236 0.060 0.344 0.774 0.425 0.093 0.070 0.039 0.330 0.258 Ring heavy atoms (rms) CHARMm TRIPOS (g) (h) 0.061 0.190 0.134 0.289)' 4R-Die1 (dl 0. bRemoved improper forcing planarity on the nitrogen in the ring.439 0.027 0. and (f) constrained (CONSR) dihedrals.138 0.039 0.449 0.931 0.141 0.307 1.144 0.058 0.438 (0.199 0.480 0.429 0.351 0.090 0.120 0.055 0.148 0.072 0.050 0.033 - 0.148 0.023 0.573 0.064 0.165 0.393 0.127 0.093 0.336 0.637 0.250 0.569 0.145 0.277 0.074 A total of 49 molecules were taken from the Cambridge Structural D a t a b a ~ e .080 0.376 0.125 0. for some of these flexible molecules. f . e. interactions between symmetry generated molecules in the crystal should be closely simulated. CHARMm uses a value intermediate between these values that gives the bond lengths shown in Table 111. the effect of changing the dielectric from C = 1 to C = 4R is shown. whereas only the heavy atoms in ring fragments were taken for the results in columns g-h. for example ABTOET does not change. It would appear that in some cases. A further study (column e) was carried out to test the effect of crystal packing on changes in molecular conformation. Those results in which MM3 is better than CHARMm are most often those molecules where dihedral-stretch or dihedral-bend terms are responsible for the deviation. the electrostatic field of the molecule is responsible for major conformational changes. The C-C bond in the ethyl-halides is also modeled as an average of observed structures. the dielectric constant was taken to be 1. The experimentally observed bond length variations. such as nonbonded interactions that may move groups relative to the positions found in the crystal structure.” This effect can be described by the observation that as more fluorines are added to a carbon the C-F bond becomes significantly shorter and the C-C bond lengths vary with different fluorine occupancy.4” from MM3).466 rms deviation. are not reproduced in these calculations. going from a value of 1. During the CHARMm calculations.450 to 0. it will change its conformation to optimize its intramolecular forces. These cross-terms are missing in CHARMm.45 A) in column a .0 for columns a. It is expected that by holding the crystal lattice at the measured values for the molecule of interest. Other structures showed only slight or no improvement in rms deviation.’ Cambridge Crystallographic Data Base Organic Molecules A set of 49 organic molecules were taken from the Cambridge Crystallographic Data Baseg and compared to simulated results (Table IV). The electrostatic driving force for conformational change in the vacuum is neutralized by other polar groups in close contact when the molecule is in the crystal environment.2. Each energy minimized structure from CHARMm was compared to the experimental structure using standard rms comparison techniques available in the molecular similarity module of QUANTA 3. In every case where unusually large rms values were found. the differences were a result of changes in dihedral angles and represent conformational rather than geometric variations. it is clear that when we remove the molecule from the crystal environment. In some cases the molecules in Table IV show large rms deviations (maximum value = 1.326 A in CF3-CF3 (see Table I11 for references on fluorine compounds). These molecules may have other reasons for the deviation. Those molecules with significant dihedral angle changes were chosen for further study to elucidate the energetic reasons for such changes. and a distance dependent form. The DREIDING3 results are given in column c. Future comparisons will show differences between gas phase electron diffraction studies and calculated dynamically averaged structures at experimentally equivalent temperatures. rms deviations from the vacuum study of just those molecules listed in column e are shown below the average . The rms deviations (see column e) of the molecules studied in the crystal lattice are much smaller than the deviations found in the free vacuum calculations (column a) and this is in agreement with the proposal that the complete crystal environment should be simulated to retain those conformations observed from the X-ray study. However.399 A for CH2F-CH3 to 1. Those molecules with significant rms deviations in the vacuum calculation (column a) were further analyzed by building the observed crystal structures in QUANTAKHARMm. All heavy atoms were used in the comparisons in columns a-f. From structural results it has been observed that the C-F bond length goes from 1. Molecule names are given in Table V and figures of the structures may be found in references 2 and 3 where this set was also studied. while in others conformational changes occur in the vacuum calculations but are not a direct result of electrostatic forces. 6. This method restricts the distances between symmetry related molecules in the crystal to be nearly constant. but are included in MM3. indicating significant variations from the X-ray structure. In column d.2”. and g. which occur as a function of number of fluorine or other halogen atoms attached.CHARMM FORCE FIELD 897 3. This level of precision is as one should expect since the experimental results (gas phase) are average measures of a dynamic molecular situation while the energy minimized structures are static minima. In the case of the fluorine compounds we have taken a middle ground by not attempting to add special functions to reproduce the “fluorine effect. Some structures such as ACBUOL show significant improvements. The Tripos2 results are reported in columns b and h. C = 4R was used and is given in column d. These studies will be described elsewhere in more detail. as measures of the ability of the force field to simulate real molecules in their dynamic states.and energy minimizing the internal coordinates of the molecules while holding the crystal lattice constants fixed. 4-Tetra-O-acetyl-alpha-~-arabinopyranose 0- Acetamidobenzamide 2-(2-Hydroxy-3-isopropylamino-propoxy)-5-butyrylamino-acetophenone hydrochloride Aeycta ctlirn Alpha-5-acetyl-2'-deoxyuridine 1.3.4.)-(1R.6-dihydro-4H-l. 10-alpha-estran-3-one 9-beta.6. again indicating that the internal coordinates are well represented.8-0-benzylidene-4-deoxy-2-ethylenedithio-~-talo-2-octulosonate 16-Alpha.4-seco-4.4.4-dimethyl-5.4. to show equivalent comparisons between the same set of molecules.)-(E)-N-acetyl-piperidine-2-carboxylic acid 2l-Acetoxy-l7-alpha-hydroxy-pregn-4-ene-3.3-diazabicyclo(3.3.l-dioxide 3'-0-Acetyladenosine 2-Acetylaminofluorene Aeaiie ctnld 1.6-tri-O-acetyl-2-(N-acetylacetamido)-2.R)-N-Acetyl-S-(2-nitro1-phenylethyl)-L-cysteine (5S)-Acetonyl-penem-3-carboxylate ( + .8a-hexahydronaphthalene (Acetoxyethy1)-trimethylammonium hydrogen ( + -)-tartrate Acetamidinium chloride 2-Acetyl-indan1.3-Di-O-acetyl1.10-cyclo-3. the rms differences between CHARMm (column g) and experimental results are very small.lH-2-benzo1-0x0- pyran-4-yZ)-propanamide hydrobromide dihydrate 2-Acetoxy-3-methylbenzoic acid NAcetyl-DL-methionine-diethylamide N .7-anhydro-6.2. 17-butanomorphinan-3-ol 3-Dimethylamino-4.4-di0-acetyl-2.898 MOMANY AND RONE n Table V .3. 6-anhydro-alpha-D-altopyranoside 2-(Acety1oxy)-benzoic acid (10s)-17-Beta-acetoxy-3. 5. To confirm that the rms deviations described above arise from conformational changes and not from incorrect geometry. the Tripos calculations reported were carried out without electrostatic in- . where the effect of partial charges upon conformation would be negligible.4.4a.2.2R.17-one N.11. Names of molecules as they appear i Table IV. In the case of ring fragments. Comparing the CHARMm results with the other reported force fields is not straightforward for two major reasons. First.9( s r d e --one 11 . N-Diacetyl-3-methylthiobenzidine 7-Thia1.6.4a. Beta-DL-arabinose Ethyl-3.6-anhydro-beta-~-galactopyranose L-N-Acetylhistidine monohydrate trans-4a-Acetoxy-8a-chloro-l.2-Dihydro-acenaphthylene cis1-Acetamido-2-fluoro-1-phenylcyclohexane NAcetyl-furanomycin Methyl-2.4-dione p -A e o o u d n cttliie N-Acetyl-L-tyrosine Beta-5-acetyl-2'-deoxyuridine ( .3R)-3-Acetyl-2-vinylcyclohexane-l-oL ( .t a i n1 )e Alpha-acetylthio-5-alpha-androstan.3-dione D L -. The rms deviations from this study compared to experimental results are shown in column f.5.20-trione Methyl-3. the variable dihedral angles (nonrings) were constrained to the experimental values and the geometry (bond lengths and angles) optimized. 0-Diacetyl-4-hydroxy-nornantenine Acenaphthylene--carboxylicacid 1 (R.4.8. Lbl ae AAXTHP ABAXES ABINOROB ABTOET ABBUMOlO ABZTCX ACADOS ACAFLR ACANILO1 ACARAP ACBNZAOl ACBUOL ACCITRlO ACDXUR ACENAP03 ACFPCH ACFUCN ACGLSP ACGLUAll ACHGAL ACHISTBO ACHNAPlO ACHTARlO ACIMDC ACINDN ACINST ACKYNU ACMBPN ACMEBZ ACMTDE ACNORT ACNPAClO ACNPEC ACPENClO ACPPCA ACPRETO3 ACPYNS ACSALAOl ACSESOl ACTAND ACTHBZ ACTHCP ACTOLD ACTYSN ACURID ACVCHO ACXMOL ACXMPR ADENOSlO - Name 1.5.6-Tetra-O-acetyl-3-chloro-2-C-(chloromethyl)-epi-inositol ~ NAcetyl-kynurenine 2-Amino-N-(3-dichloromethyl-3.5-tetra0-acetyl-alpha-D-galactoseptanoside N-acetyl-alpha-o-glucosamine 2. These results clearly confirm that the geometry of the molecules is calculated to be in excellent agreement with experiment.)-1-0-Acetyl-xylomollin 3R-(1'(S)-Aminocarboxymethyl)-2-pyrrolidone-5-(S)-carboxylic acid + Adenosine deviations in columns a-c.7-hexahydr0-5.5-benzothiazocin-6-one1.O)octa-2.8-trihydroxy-3-methyl.3-dideoxyalpha-~-threo-hex-2-enopyranose 2-alpha-bromo-l7-beta-acetoxy-9-methyl-5-alpha. . V. Klimkowski. (c) V. S.A. Momany. and E. Soc. Struct. thus they eliminated important polar contributions which affect the conformations of many of the molecules studied here. Klimkowski. 187 (1983). G. Soc. Chem. 2. atom types. and C. B. 4665 (1991). Momany. (b) L.D. Brooks.J. Because of this. (h) H. J. Phys. Siam. L. (e) L.. Prog. 4768 (1990). using the gradient cutoff conditions used in the Tripos article2 with no electrostatic contributions. J. Klimkowski. Basch and W. 4768 (1990). 2331 (1979). Higgs. Hosoya. Phys. J. Brice. (a) J. F.. Momany. and F. H. F. Scarsdale.R. Am. Org. S. many different atom type combinations for angles. Schafer. highly flexible. Mol.N. E. Stevens. Mayo. J.. Newton.722 (1988).J.G. 8.. J . Chem.S.A..L. Schafer. 105.. Jasien and W. Chem. 261 (1987). and C. in preparation. Momany. Osawa. Soc.D. Struct.. 654 (1990).A. Schafer. Comp. while dihedral angles change in response to electrostatic and nonbonded interactions much more slowly.A. and L.R. (b) B. Van Alsenoy. Lovas. and nonbonded parameters are added. (g) C. Tom Halgren. Klimkowski. (b) K. Chem. Jonsson. C. one must consider the effect of premature termination of the minimization process. Chem. July 1991. Van Opdenbosch.L.. Phys. Momany and R. Motherwell.D. J. 143 (1985). have refined parameters and are easily accessible. 1439 (1982).Q. Siam. Polygen Corporation. The large number of parameters routinely available through QUANTA 3. Goto. 10. (d) F. bonding parameters.J. M. L. conformational energy differences. Schafer. Chenz. F. F. Storer. Schafer. The diversity of types of molecules that can be studied using QUANTAKHARMm is impressive and becoming greater every day as new partial atomic charges.H.J. Chern. Bellard.D. Merck Sharpe and Dohm. 124.. C. During minimization bonds and angles respond to harmonic forces very quickly. Wallqvist.. 23. Soc. 3271 (1986). Olafson. That is. and Supplement to Parameter Handbook.. Am. and N. Karlstrom.CHARMM FORCE FIELD 899 teractions'. D. Mol.G. (d) F. and D. H. Severance. Allen. Klimkowski. Struct. 7. in press. Schafer. 112. 112. B . in press.H. R. P. etc.E. 569 (1985).2. Tanabe.* . 84. 5. A m .2 allows the study of new compounds with complex microenvironments. Second. Simulations of cyclic-polypeptides and comparisons with X-ray results have been recently published6 and are not included here.P. Chem. (i) W.. Ewbank. V. Tsuzuki. dihedrals. Momany. Jorgensen and C . Chem. T. Momany. 275 (1990).D. when we studied single molecules in vacuum. Comp. V.D. J. This allows broad molecular variability without the difficulties incurred when one defaults to generic parameters. Lett. Zozom. Cramer 111.. Jorgensen and D. Soc. Mol. and L. Gillies. 105. 3777 (1983). Reynolds. 18. C. and geometries.. Data presented here show that highly accurate geometries and conformations are obtained using this force field.E. and V.. Watson. Personal communication. H. Soc. Bruccoleri. June 1990. A. J. S.L.G. and L. the result of which may be unrealistic. Suenram. 6. Linse. 4.376 for the set of molecules listed in column e of Table IV. States. W e consider that a gradient over all atoms of at least 1 x is necessary in order that the soft variables (dihedral angles) have had ample opportunity to vary from the starting structure.D. 94. Hummelink-Peters. References 1. J. 1256 (1979). but is in reasonable agreement with both the Tripos and DREIDING values in columns b and c (values in parentheses) of Table IV. These studies will be reported in detail elsewhere. J. J. These variables are notoriously slow in reaching a minimum because of the very flat surface around their energy wells. 4. Chem. Am. S. (e) W. Chem. K.0. Newton. Scarsdale. Huang. Struct. Rone. Comp.G. Stevens.R. Phys. Rogers. CONCLUSIONS The CHARMm force field has been shown to be an all purpose. 1.. F. M. and M. 7903 (1990). R. S. Lowe. 2335 (1984).. Fraser.. Smith. This is clearly less than the 0. Clark. (a) B. J .Q. QUANTA Release 3. S. 113. Phys. Severance. 169. Chem. Am.T. Van Alsenoy. Chuman. F. B35. B. Renugopalakrishnan. Doubleday. Parameter Handbook. For example. Swaminathan. 982 (1989). V. and J.J.D. A. Olafson. Van Alsenoy. Biopolymers..P. The geometries and conformations produced using CHARMm on these polypeptides show extremely close correspondence to structures found experimentally from X-ray diffraction data. in press. Van Alsenoy. J . Chem. Eds. Schafer. Chem. Comparisons were made to experimental results on rotational barriers. 10. 9. (a) L. Acta Crystallog. Klimkowski. (j) S. 152. Phys. Expanding Frontiers in Polypeptide and Protein Structural Research. J. W.. 11. J. Jemmes. V.R. Waltham. J . J. well parameterized force field for small organic molecules. W.L. Schafer.A. and J . Molecular Systems Department. V. (f) F. MA.A. Am. J. QUANTA Release 3. (f) C. Biochemistry. J. 76. (c) L. Karplus. Schafer. J.802 value found for this molecule set when using the full CHARMm force field with a low value for the gradient. Schafer. and I. and V. Soc. Soc. Goddard 1 1 J . Klimkowski. 0. J. ( c )P.A. Chem. R. K. Klimkowski. Van Alsenoy. Swenson. Mol. H. P. Dykstra. A m .A.N. 1 (1968). 88.A. A. Gelin and M. we obtain an rms deviation of 0.A. J. 3438 (1983). 112.J. (b) W. L. Momany. 6. 107. incomplete minimization can lead to an artificially low rms deviation when comparing the calculated to the experimental structure. Chem.L. and W.C. Hummelink. J. 112..J. Carey. B. 8897 (1990). B. J. (a) G. C.J. Kinnard. Karplus. 7540 (1990). 11. 3. One does not expect the conformation of polar molecules to be the same in a vacuum situation as in the crystal for this very reason.. J. Am. Jorgensen and D. Cartwright. J. Chem. and B.L. L. 11. Struct. T. 144. J. 41. Phys.A. and C. Am. Allinger. Teixeira-Dias. S. Jpn. Zanjanchi. Van Schaik. J. H. N. Higginbotham and L. Struct. Fausto. N. and M. Lane.S. P. K. Morino. 31 (1980).F..D. Kimura. A. Struct. 111.. Li.L. and L.. 27.. Struct. Soc. J . 31.R. J. 102. K. . Struct. 26.. 1291 (1972). Struct. Soc. 18. Jpn.R. Soc. and H. Van der Enden. 16. 56. Allinger. 67 (1971).K.. J . Jones. Molecular Constants.M. 151 (1959). 35. Chem. Soc. J . 51. 18. J. 36. 465 (1981). Chem. R.L. 40. Rahman. T. and P. Jpn. 116. Bartell.L. Mol. Fausto. 34. Chem. 39 (1976). Comp. Mol. Group 11. Ewbank. Pankrushev. 2. Chem. Kirsch. Chem.S. 14.H. Siam. 131 (1984).L. (b) B. N. MOMANY AND RONE 29. 9. 99 (1984). Chem. Ed. 23. Kveseth. B. vol. A34. Phys. 231 (1976). J. Kimura. 118. 42. Mol. T. Chem.. 1971 (1975). Chem. Mot. Bartell and H. Tetrahedron. Mol.2189 .S. 2564 (1977).. R.H. A. J. Iijima. Tai.L. K.C..A. J . T.203 (1981). Iijima. 50. and J-H Lii. and L.. J. Van Alsenoy. L. G. Yan. J. S.. J. Hellwege..J. and Y. Mot. Struct. L.H. 20. 39. 42... Mol. Struct. Comp. and M. L. (b) J. A. M. Chem. 1726 (1966). subvol. N. Acta Chem.. 44. Haugen and M. Bradford. 113 (1977). D. J. 215 (1986). 163 (1973). Bull. J. Teixeira-Dias. Teixeira-Dias and R. Comp. and D.... (1980). 137 (1987). J . Chem. Winnewisser. J. 11. 6 . and M. 125. J. Y.Y. 331 (1985). K. D. Vilkov. Chem. Khim.O. Scand. Chem. 1982. Kimura and M. J.. 159.. 343 (1957). 112. 45. L. Fernholt.. Soc. Kuchitsu. Konaka. (a) R. K. and J. J. 33. M. 37. (a) B. (a) 0. Schafer. and G. C. Kohl. Spectrosc. 8307 (1990). Klimkowski. Murayama.L. MoZ. N. Fukuyama. 351 (1984). 868 (1990). Bartell. L. Geise. 1270 (1983). L. Grigoras and T.. 25. 8551. 17. Lide. Allinger. J. Springer-Verlag. 22. 1131 (1965). LandoltBoerstein. 42. Kuchitsu. S. Bull.I. 15. A m . H.C. H. Struct. K. J .L. 27.. Jones.M. J. Takemura. Geise. I.C. Konaka. J.M. .S. Shichkiv. 32.R. S. Jpn. Scand. 144. Kuchitsu. M. J. 21. 43..900 12. J . Strukt. Meyer and N. 13. 30. Van Nuffel. Heenan and L. Schafer. Chem. Kimura.. Mol.. Am. A m . J. 24. Allinger. Kimura.. Tsuboyama. Yavari. Actu Chem. Allinger. W.J.. M. Yuh. Mol. Bartell. Soc.L. Bartell and W. Baegley.V. 71. 31. 25 (1988). Bradford. Traetteberg. Seip. 30.K.. 78. 199 (1986). 42. Iijima and M. 848 (1990). 46. Soc. J. N.M..A. Struct. Derissen. R. 24. 20.H. and E. 8566.J. Bull. Soc. Schmitz and N. 86. Sadova.V. Hemelrijk. Yan.. Struct. C. Mol. 185 (1977). Fausto and J . Van Ophensden. Mol. Chem. Kambara. Mol. 19. J. M. 1.. and K.C. and L.. S. J. Van Alsenoy. Seip.F. Higginbotham. Mol. 8293 (1990). 25 (1983). Mol. H. 215 (1979). 45. Chem. Van den Enden.. Oyanagi and K. M. J. Seller. 7.C. J. 2237 (1978). Chem. Chem. Allinger. Mol. Struct.. and M. F. 851 (1965). L. Tsuchiya. Lees. 14. Bull. and Y.L. 8576 (1989). 61. J . Van Zoeren. 112. Comp. Fitzwater.P. Phys. Struct.S. J. Spectrosc. 37. Kubo. L. Phys. Phys. Van Eijck.. Tamagava.L. Allinger. 31. 2159 (1969). Gelly. 463 (1968). and J-Huei Lii. V. and P. A. 28. 38. J. Beagley. J. Schafer.F.O. Carey. (b) W. New York. 38. Struct. Bastiansen. and L. J. K. R.
Report "Validation of the general purpose QUANTA ®3.2CHARMm® force field"