and Dynamics Calculations* and Martin Karplus Department vf Chemistry, Haruard U Received 6 ly 1882; accepted 14 Detober 198 mal mode or molecular dynamic: ‘sample run are included. 1, INTRODUCTION Invtacent years, the usefulness of empirical en ergy functions for investigating the physical properties of a wide variety of molecules has been domonstrated.'-® Studies range from those eon cerned with the structures, energies, and vibra tional frequencies of small molecules," Ynrough those dealing with Monte Carlo and molecular dynamics simulations of pure liquids and solu tions," to the analysis of the conformational energies and fluctuations of large molecules, such 48 proteins and nucleie acids in vacuum, in solu. tion, and in a erystal environment.2* It is de- sirable to have available a general and flexibie computer program which ean efficiently handle all aspects of suck computations with empirical en- engy fonetions forthe various systems of interest, In tis article we describe the design of such @ program (CHARM) and give details concerning ite implementation, It is possible with the present version of the program to model build the strue ~* Supported in part by grants frm the Nation Science Founsdation and the Notional insiutes of Health ‘TNational Science Foundation Postdoctoral Fellow National Science Foundation. PredoctoralPellow 977-1980, ‘National Institutes of Health Postdoctoral Fellow rset address: Nave Laboratory af Chemica Pyas, alitornia Institute of Technology, Pasadena, Californie sues “ 1 Support by the National Instites of Health Med- inal Schnee Training Program, grent No. GMO-7753. Journal of Computational Chemistry, Vol. 4, No. 2, 187-2 ‘© 188% by John Wiley & Sons, Ine Bernard R. Brooks,’ Robert E. Bruccoteri Barry D. Olafson,’ David 1. States,? cHaRMa (Chemistry at HARvard Macromolecular Mechanic which uses etpirical energy functions to model macromolecular systems. The program can read or ‘model build structures, energy minimize them by first: or second-derivative techniques, perform a nor mulation, and analyze the structueal, equilibrium, and dynamic properties determined in these calculations. The operations that CHARMM can perform are described, and some implementation details are given. A set of parameters for the empirical energy function and at CHARMM: A Program for Macromolecular Energy, Minimization, ‘Swaminathan, Cambridge, Massachusetts 02138 is a highly flexible computer program ‘care of interest, energy minimize that structuse by first- and second-derivative techniques, perform 4 normal mode of molecular dynamics simulation, and analyze the structural, equilibrium, and dy- namie properties determined in thexe calculations. Currently, CHARMM can treat isolated molecula molecules in solutions, and molecules in erystalline solids, The information for proteins, nucleic acids, prosthetic groups (eg. heme groups), and sub- strates are available, and the extension to other molecules is straightforward. A wide range of analysis is possible, inchoding statie structure and energy analysis, structure and energy compafisor time series, correlation functions and statistical properties of molecular dynamic trajectories, and interfaces to computer graphies programs. In developing the program, we have concen: Uuates on three aspeets ofits design that lead to an efficient calculational system. Fisst, the program should complete a specific task in as short a time as possible; second, we have attempted to simplify the input necessary to obtain any desired results; and third, we have tried to make it easy to under stand and modify the program. Iti these aspects of the program which have concerned us most and which have led to some original aspects in its im- plementation ‘The arganization of the program is presented in Figure I. Table I gives a brief description of each ‘of the boxes in the figure. The support routines indicated in the figure provide the programming (1983) ‘cee o192.86851/83/0 0187-91804.10Brooks etal Figure |. Organization of CHARM. The top half of the figure shows the major ‘groupings of commands under CHARM, The bottom hall outliees the support ‘subroutines which provide the programming enviranment for the implementation of ‘CHARM, Table I gives a verbal descripcinn of the boxes in the figore, of HARM (see Pig. 1), eeHaram environment for the implementavion of the com mands in CHARMM. In this article, we describe the energy expression in See. IT and the generation of data structures required to compute the energy in Sec. IIE. This order is followed because the energy expression provides the rationale for the generation process. The generation af unknown coordinates and the manipulation of conformations is deseribed in See. TV, InSee. V, we outline the energy manipulations required for energy minimization, normal mode analysis, and dynamics. The analysis facility is discussed in Sec. VI. In Sec. VIT, we treat specitie programming aspecls af CHARMM that made it possible to implement the desired design features of CHARMM, Section VIII presents the discussion and conclusions. Appendix 1 gives a set of pa- rameters for the explicit hydrogen representation of molecular structure, Appendix 2 gives a sample CHARMM run on a small molecule which demon. strates many of the program features. A table of execution times for a few representative tasks is siven in Appendix 3. A number of programs with some similar fea tures have been deseribed 3“ By not comparing these programs with the one presenter! here, we do not mean te imply that our implementation is preferable, but rather to limit the discussion to the present program and to demonstrate how its de sign features have made it a useful research tool The closest in conception and formulation to the present program is the program AMBER! which is based, as is CHARMM, on the macromolecular Program developed at Harvard by Bruce Gelin and Martin Karplus #4 I, THE EMPIRICAL EN! FUN In this section, we describe the empirical energy function that is used in CHARMM. A. Atomic Representation ‘The fundamental unit used in CHARM is the atom, An atom is considered to be a charged point ‘mass with no directional properties and wit internal degrees of freedom. Masses are expressed, in atomic mass units. Each atom of a real system could correspond to a single atom in CHARMM, but for targe systems some (or all) hydrogens are combined with neighboring heavy atoms to which they are bound. This combining of atoms is re 189 ferred to as the “extended-atom representation.” For the peptide group and amino acid side chains, it has been shown that the extended-atom dihedral angle potentials can essentially reproduce those obtained from an all-atom representation. Some advantages and disadvantages of the ex- tensied-atom approximation are presented. Advoataxes: (1) ‘The most important advantage of extended atoms is that they significantly reduce the size fof most problems, since roughly helf of the ‘atoms in biological or other organic macro- molecules are hydrogens. This results in fewer nonbonded interactions and internal degrees of freedom (2) Motions invotving heavy atoms are separated from hydrogen stretching motions. The large gap in the IR spectra between these types of motions implies that. removing one type should have # small eflect on the other? (3) Larger dynamies integration step sizes ean be used by not including hydrogens since the small mass of hydrogen atoms requires @ smaller time step for acrurate integration. (4) Hydrogen positions are usually not available from x-ray crystallography and must be gen- erated from the heavy atom coordinates Disadvant (1) Itis difficult to represent accurately hydrogen bonding, since the position of a hydrogen atom has a large effect on hydragen bond strength. When only heavy atom positions are included, a crude approximation to the hydrogen bond must be introduced. (2) Dipole and quadrupole moments are lost combining hydrogens into extended atoms; 4. the loss of a dipole in a NH bond can adversely affect its interaction with other nearby groups, [3) There isa (oss of steric effects to hydrogens, since an extended atom is always spherical, Hydrogen coordinates are necessary for some types of analysis (e.g., proton and “C-NMR phenomena). in ‘The extended atom representation has been shown to provide a satisfactory representation of the internal vibrations and bulk properties of small molecules including simple peptides.239484849 Jp the past, when computational considerations posed severe limitations, the extended-atom rep-