Chemical compounds have a biological activity on animal genes; Mutagenicity and Carcinogenicity are examples of this activity. Both activities are similar except that mutagen compounds induce a heritable change in cells, while carcinogen compounds induce an unregulated growth process in cells. The relation between Mutagenicity and Carcinogenicity is not proved quantitatively yet. In this article, the relation between both activities is discovered based on the machine learning methodology. Feature selection techniques are applied to provide a well defined analysis of the highest discriminating descriptors of the mutagenic or carcinogenic compounds. Molecular charge appears to be a discriminating factor for Carcinogenicity, while Mutagenicity is characterized more by the branching and aromaticity/aliphaticity of the compounds. Electronegativity and Lipophilicity appears to be common factors for both activities. Further analysis and visualization are applied based on rough set and formal concept analysis to check the correlation among these descriptors and the ranges required for each descriptor.