RESUMEN
The efficiency of estimating the carcinogenic activity of chemical substances was compared for five short-term tests and structure-activity relationships (SAR) analysis. The sample included 84 substances with known biological testing results obtained by the Ames test, bacterial SOS chromotest (SOS), chromosome aberration (CA) cytogenetic test, sister chromatid exchange (SHE) test, and gene mutation (GM) test with mammalian cells in vitro and by carcinogenicity assays in rodents in vivo. Structural descriptors were selected using an original database, which included the structural formulas of substances with known carcinogenic activity in rodents. Original software was created to generate and select the descriptors that statistically coincided with carcinogenic activity. The descriptors that were associated exclusively with carcinogenic substances from the database and the tests that produced positive results exclusively with carcinogens were used to evaluate the carcinogenic activity of the substances. A combination of three short-term tests (Ames, SOS, and CA tests) and the SAR model proved to identify carcinogenicity for more than 60% of carcinogens.