The ever-growing number of large available chemical and biochemical databases present new challenges, which often demand an interaction of computational and chemical methodologies. Relevant chemical and biological knowledge is embedded in databases that can only be revealed by computer approaches. Machine learning algorithms can identify relationships within data sets, or inconsistencies, and can build models. Such approaches are extremely useful for the application of available data to the resolution of new problems, for the maintenance of databases, or for the formulation of new hypotheses, adding value to existent experimental data. The Chemoinformatics Lab of FCT/UNL has devoted to such endeavors since 2001.