Empirical Explorations of SYNCHEM
H. L. Gelernter 1,
A. F. Sanders 2,
D. L. Larsen 3,
K. K. Agarwal 4,
R. H. Boivie 5,
G. A. Spritzer 6, and
J. E. Searleman 6
1 Professor of computer science at the State University of New York, Stony Brook 11794
2 Assistant professor of computer science at Wright State University, Dayton, Ohio 45431
3 Lederle postdoctoral fellow at Stony Brook
4 Assistant professor of computer science at Wayne State University, Detroit, Michigan 48202
5 Member of the research staff at Bell Laboratories, Holmdel, New Jersey 07733
6 Graduate students in computer science at Stony Brook
During the past several years, a substantial body of experience has accumulated in the use of SYNCHEM, a large-scale program which is able to discover synthesis routes for relatively complex organic structures without on-line guidance on the part of its chemist user. These results indicate that the approach to computer-directed organic synthesis route discovery embodied in the program has been valid and reasonable, and that SYNCHEM is likely to be fruitful from the point of view of its intended users as well as for our research objectives in artificial intelligence. The experiments have revealed a number of insufficiencies in the program as well. Most of these are rectified in SYNCHEM2, a revised version of the program which includes, among other improvements, a more highly developed synthesis search algorithm and the routine consideration of stereochemistry.
