Brutlag, D. L. (1988). In R. R. Colwell (Ed.), Biomolecular Data: A Resource in Transition (pp. 185-188.). Oxford: Oxford university Press.
Department of Biochemistry, Stanford University Medical Center, California 94305.
A knowledge base capable of predicting the activity of DNA polymerase I under a wide variety of conditions was developed using classical deduction rules together with a system for truth maintenance. These rule-based methods have the intrinsic property of being able to explain conclusions both in the form of English sentences and in the form of graphic representation of the flow of inference. Variables such as the presence of salts, nucleotides, temperature, ionic strength, pH, etc. are all represented graphically as active images of thermometers, gauges and switches. Properties of the physiological conditions and substrates can be changed by using a mouse pointing device. In response to changes of any parameter, a series of rules defining the specificity and reactivity of enzymes are automatically invoked and specific conclusions are made. The methods for specifying the actions of enzymes is well defined, allowing ready simulation of other enzymes. These simulations can be coupled together to generate a discrete event simulation of multiple steps in a metabolic pathway.
The power of rule-based systems are multiple. One can represent the knowledge of the metabolism at many levels simultaneously. In some cases, the actual mechanism is known and intermediates in the reaction can be represented. In other cases only the substrates and products are known and these can be related by appropriate rules. The speed of rule-based inferences is much faster than that of mathematical models making interactive simulation possible. Changes and addition in knowledge about an enzyme can be readily accommodated as rules can be added or removed without regard to their order.
[Back to Doug] [Address] [Academics] [Honors] [Publications] [Presentations] [Public Service]