A STUDY OF QSAR OF NITRATED POLYCYCLIC AROMATIC-HYDROCARBONS USING A KNOWLEDGE-BASED SYSTEM APPROACH

A knowledge-based system approach has been suggested to study the relationships between molecular structures and mutagenicities of nitrated polycyclic aromatic hydrocarbons. A series of feature molecular structure descriptors on which the mutagenicities depend closely have been selected by using the classic pattern recognition method. Then the feature descriptors are input in an expert system ''shell'' EXSYS, to build a simple mutagenicity classification model based on decision rules. Basing upon four classes of activities, the direct-acting mutagenicities in Salmonella Typhimurium TA98 of 80 nitrated PHAs are distinguished one by one, and the correct recognition rate can reach up to 0.91. This result is better than what was obtained by only using the pattern recognition method based upon two classes.