Elsevier Science Home
Computer Physics Communications Program Library
Full text online from Science Direct
Programs in Physics & Physical Chemistry
CPC Home

[Licence| Download | New Version Template] abjg_v1_0.gz(29 Kbytes)
Manuscript Title: MOLSIMIL-88: molecular similarity calculations using a CNDO-like approximation.
Authors: R. Carbo, B. Calabuig
Program title: MOLSIMIL-88
Catalogue identifier: ABJG_v1_0
Distribution format: gz
Journal reference: Comput. Phys. Commun. 55(1989)117
Programming language: Fortran.
Computer: IBM PC XT/AT 286 - 386.
Operating system: PC (MS) DOS VERSION 3.2 & 3.3, VAX VMS VERSION 4.
RAM: 256K words
Word size: 16
Peripherals: disc.
Keywords: Mo lcao theory, CNDO approximation, Sto ns basis functions, Gross atomic populations, First order density Function, Molecular similarity Measures, Molecular correlation Coefficient, Artificial intelligence, Pattern recognition, Principal components Analysis, Kruskal tree, Nearest neighbourhood Graphs, Structure, Properties relationships.
Classification: 16.1.

Nature of problem:
Molecular Similarity Measures are performed between two molecular structures throughout their electronic density distribution approximated using a Complete Neglect of Differential Overlap (CNDO). The parameters chosen here as Similarity Measures are a Correlation Coefficient and a related Euclidean Distance. Within a family of molecules the pairwise similarity measures build up a positive definite matrix, which can be used to construct trees and graphs or to analyze the correlation matrix principal components. The molecular family can be visualized in various manners as a whole, and the resultant pictures can be used to find graphically molecular subsets or groups, as well as to blow up Qualitative Structure - Properties Relationships.

Solution method:
The main program calculations are performed according to the general scheme of reference. Some features, related to the molecular similarity measure optimization, and other details, not commented on before nor published, but discussed in the forthcoming Long Write-up, are also introduced into the present code.

The present program, designed to be transferable and compatible as far as possible, has all dimensions adjustable. Thus, depending on the CPU RAM available in a PC-like machine, the number of molecules to be compared can be reduced or extended by changing one numeric constant only. The present version has a capacity to handle 50 molecules, but it can be augmented until the PC DOS restriction 0f 640 Kb addressing limit is reached. In a VAX computing environment the parameter file has been maintained as in the PC version, but memory restrictions are not so harsh.