Molecule, design automatic synthesis

The automatic synthesis of molecules, described in the previous section, attempts to generate feasible candidates without resorting to the detailed, fragmented, often informal, but nevertheless sound and valuable knowledge possessed by the human designer. The search for new molecules is carried out efficiently, but it is entirely built on a limited amount of knowledge, which is represented by the sets of constraints on (1) physical properties, (2) the feasibility of molecular structures, and (3) chemical stability and other chemical properties of the resulting molecules. 

The generate-and-test search paradigm, used by the automatic design algorithm for the synthesis of molecules, is composed of two modules. The first module, the generator, enumerates candidate molecules. The second, the tester, evaluates each molecule, estimating its physical properties and checking for structural feasibility, and either accepts or rejects it. We... 

Additionally, since our design strategy involves the transformation of one molecule into another it was necessary to develop a means to do this automatically if it were to be efficient. Performing the chemical transformations with a text editor incurs the risk of generating nonsense chemical structures. Thus for this purpose we needed to describe to a computer program how to transform the database molecules to the compounds suggested for synthesis in such a way that only chemically valid structures would be produced. 

---