Predicting carcinogenicity with short-term tests: biological models and operational approaches.

This paper illustrates the basic biological models that have had an influence on the development and deployment of short-term assays for carcinogens. A tentative description of such a conceptual framework is provided with the use of methods for the quantitative analysis of information. This picture is contrasted with the results of a series of statistical mathematical analyses carried out on real genotoxicity data bases. The resulting evidence indicates that the biological models derived from basic laboratory research are not adequate or sufficient to interpret the operational performances of short-term assays. Obviously these models are of great importance, but, this observed discrepancy suggests the need for articulated approaches, with particular emphasis on analytical tools able to interpret the complexity of these systems. In particular, the multivariate data analysis methods are indicated as suitable for the description of large and complex bodies of data, and the identification of typologies and regularities. A number of applications of such methods are also presented. The conclusion is that biological models and operational approaches have to interact on the same level. The operational approach should indicate whether the conceptual framework of the experimenter is adequate to describe the experimental situation and possibly point to new relationships and trends, as well as to practical solutions.