MOLECULAR MECHANICS AND THE CAMSEQ PROCESSOR R. POTENZONE, JR.*, ELIZABETH CAVKCHI*, H. J. R. WEINTRAUB? and A. J. HOPFINGER*$ (Received 14 January 1977) A&&-The CAMSEQ processor used to carry out molecular mechanics as well as other molecular structure calculations is described in detail. General comments concerning potential energy functions, exploration of conformational space, and optimizing, with respect to size and speed, structure and coding of various subroutines of a molecular mechanics package are inserted at appropriate positions in the CAMSEQ description. Size, speed and reliability benchmarks for different versions of CAMSEQ on different computers are also provided. INTRODUC’ ITON The goal in the development of a molecular mechanics program is the simultaneous maximization of the speed and accuracy of the structural calculation. Beyond this basic endpoint different investigators have sought in- dividual goals consistent with their particular interests in the structural chemistry of molecules. Consequently, different criteria of achieving computational speed and structural accuracy have emerged. In addition, most of these individual criteria have been influenced, or at least tempered, by the user’s ‘philosophy’ of what is to be expected from a structural calculation. For example, some workers are willing to sacrifice speed for simple input (from the user’s point of view), while others sacrifice accuracy for enhanced speed by adopting ‘sim- ple’ potential energy functions. We do not wish to express our opinion as to what constitutes a ‘good’ molecular mechanics program. We would like, however, to discuss the theory and function of a family of software packages, code named CAMSEQ, which we are employing in our structural studies. THE CAMBQ PROCFBSOR There are presently three versions of CAMSEQ, all written in Fortran, in use; two batch rnilde processors (at Purdue and at CWRU) and an inrcractive version available on the NIH PDP-10 network. We will restrict our discussion to the specific features of the PDF-10 version because, it is (a) the most sophisticated package, (b) the most accessible package to the potential user, and, (c) able to ‘talk’ to a host of complementary pack- ages at the NIH providing it great utility. The or- ganizational structure of PDP-10 CAMSEQ is shown in Fii. 1. This figure will be used for reference in the following discussion. The description of the system is broken down into the set of partitions below. Within each partition we discuss what we believe might be an optimum means of achieving a computational task and compare that to the method currently employed in CAM- SEQ. *Department of Macromolecular Science, Case Western Reserve University, Cleveland, OH 44106, U.S.A. tlkpartment of Medicinal Chemistry and Pharmacognocy, School of Pharmacy and Pharmacol. Sciences, Purdue University, West Lafayette, IN 47907, U.S.A. ZSloan Research Fellow. (a) Input There are two distinct sections of input: first, the molecular structure to be analyzed and, second, the control information used to design the properties of the structural analysis and the presentation of the results, i.e. output. The molecular structure is input as a connection table. Three options are currently available. The user may choose a structure directly from the NIH-XRAY files and pass it onto CAMSEQ via the file name. The user may have the graphics tablet and input the con- nection table via this route (this is only available cur- rently on a sub-version of CAMSEQ developed by Dave Pensak at DuPont). Lastly, the user may generate a connection table file. Here he has a few options; (a) only a connectivity matrix is input, (b) coordinates (either frac- tional coordinates and unit cell dimensions or Cartesian coordinates) of all atoms of the molecule and con- nectivity matrix or, (c) a partial set of coordinates, bond lengths and angles spanning the connectivity matrix. A fourth option of connection table input based upon as- sembly of molecular fragments, i.e. -CH,, -C6HS, etc. groups, is not yet operational. The formulation and presentation of the structure analysis in CAMSEQ is achieved through a series of operation cards. The most significant classification of the operations is: (i) Whether an initial valence geometry energy mini- mization is to be done on the initial 3dimensional struc- ture given to CAMSEQ as input or generated internally by the model builder section. Work is currently un- derway to allow the user to selectively refine sections of a molecule using MM1 (The valence geometry op- timizer): (ii) Whether atomic charges are to be input as part of the connectivity matrix or computed internally; (iii) The type of conformational analysis to be per- formed. This will be discussed further under the heading of The conformational analysis processor; (iv) Specific potential energy functions for the com- putation of hydrogen bond, coulombic and solvaticn energies; (v) Which connection table format option is used. (b) Preparation of the initial reference confomzation The initial and reference-state three dimensional geometry is established in this section. The user has the capability of defining reference dihedral angles about all bonds so that an unambiguous assignment can be as- cribed to all conformational bond rotations. The atomic 187