QSAR Approaches in Mutagenicity and Carcinogenicity Estimation
Romualdo Benigni, Alessandro Giuliani · 1997
Although QSAR is primarily a modern technique pioneered in the drug industry, the concept of the relationship between chemical structure and biological activity was established in the nineteenth century in the area of toxicology. Moreover, after the modem formulation of these ideas in the 1960's in the field of medicinal chemistry, it soon became apparent that the QSAR models could be used to formalize and investigate any type of biological activity, including toxicity. The frequency and the systematicness with which the QSAR approaches are used in medicinal chemistry are unparalled in toxicology. There are also great differences in the use of QSAR in the various areas of toxicology. This work presents the QSAR studies produced for the mutagenicity and carcinogenicity endpoint. As in classical QSAR applications to medicinal chemistry, various studies relative to individual classes of mutagens and carcinogens describe how the potency of the active chemicals in each class varies according to the variation of chemical structure/properties. However, in mutagenicity and carcinogenicity QSAR studies, the need for investigating the difference between active and nonactive chemicals has a primary importance, since risk assessment is concerned first with this issue, and second with the potency of the active compounds. A number of examples show how fundamentally distinct types of QSAR models may exist for the same chemical class: those that model the conditions for distinguishing between activity classes, e.g., “actives” and “inactives”, and those that model the conditions for modulating potencies among the “actives”. Moreover, the needs relative to the practice of risk assessment have motivated attempts to construct general QSAR models (e.g., for predicting chemical carcinogenicity), not tailored to congeneric series of chemicals, with the ambitious hope that these models would be valid for all kind of chemicals. The performance of these attempts in validation studies has usually been limited, and the results of the predictions for the individual chemicals have shown that the various SAR and QSAR approaches essentially acted as gross “class-identifiers”: they pointed to the presence or absence of alerting chemical functionalities, but were not able to make gradations within each potentially harmful class. The possible reasons of these limitations, together with future perspectives are discussed.