The wide availability of computer software for statistical analysis, chemical structure manipulation, and process simulation has resulted in the development of thousands of physical property estimation techniques. Many of these techniques are highly tailored, applicable only to a specific chemical family or a narrow range of state conditions. Computerized management of estimation techniques is, thus, essential today. We discuss a new class of databases, called knowledge bases, that, in addition to manipulating numeric and textual physical property data, manipulate molecular structures and physical property estimation techniques.