Neural Systems That Extract Information from Databases

Our company will develop a method of automatically programming neural-networks to extractneeded information from databases of biological and chemical information. We will pursue the creationof a framework for training a neural network to predict the toxic and carcinogenic properties ofchemicals using data stored in the Organ Toxicity Database. The neural network will be trained torecognize patterns and correlations between chemical structure/ substructures, physicochemicalproperties, and toxicologic/carcinogenic findings stored in the electronic compilation of a large subsetof NTP TR and NTP TOX documents. A system which automatically selects the pertinent data fieldsfor accurately predicting toxicity and then trains a neural network will be developed. The ability of theresulting neural network to predict the carcinogenic and toxic properties of known liver and kidneytoxicants and carcinogens will be evaluated during Phase I. If accurate prediction is achieved, we willbuild a generalized framework for extending our techniques to other types of biological and chemicaldatabases during Phase II. The value of many databases will be greatly increased by the availability ofmethods to automatically create neural networks which extract needed information for variousapplications. For example, our toxicity prediction system would reduce the exposure of the generalpublic to possibly hazardous chemicals, would allow chemical manufacturers to quickly determine thehazards of new and existing products, and would save enormous amounts of time and money currentlyspent on animal testing. Each application developed using our framework could offer similar benefits.