Computer Aided Synthesis Design at RISC-Linz: Automatic Extraction and Use of Reaction Classes, Blurock, Edward S.; J. Chem. Inf. Comp. Sci., 30, 505 (1990).
This work is based on a system, actually written in LISP, which took as input an extensive chemical reaction database (provided by Molecule Design, pioneers in reaction databases) and using a tuned maximal common subgraph algorithm, written by me, to determine the “reactive center” of the reaction, i.e. the general reactivity described by the specific reaction. This research was the forerunner of my work both in machine learning, I was “learning” to describe general reactivity from specific reactions from a database, and how molecules, represented as graphs, change their connectivity during a reaction.
Graph Theory: The heart of the technique represented molecules as colored graphs. The colors of the nodes including atomic information and information about multiple bonds and resonance. Reactions in this representation are a set of reactive graphs and a set of product graphs. What changes in the reaction is the connectivity of the nodes (and possibly valence state). The tuned maximal common subgraph algorithm was used find how the connectivity changed. Chemically this means the fundamental reaction that the reaction belongs to, with a reaction class being a set of reactions with the same reactivity. The “tuning” of the algorithm used the fact that the colored graphs (describing atomic structure) involved did not have a complicated connectivity (mostly linear, only up to four connections) and other factors using the atomic information.
Algorithms and data structures: The LISP system itself was a complete system not only for the fundamental algorithm, but also for the management of the database and the management of the algorithms involved.
Artificial Intelligence: Besides being a complete database expert system, the algorithm used could be considered a learning algorithm, i.e. finding generalities (in this case fundamental reactivity) among specific cases.
Computer-Aided Organic Synthesis: Development and Implementation of a Complete Synthetic Strategy, Blurock, Edward S.; Tetrahedron Computer Methodology, 2, 207 (1989).