The paper illustrates the chemometric strategies appropriate for extracting information from a large amount of biological data regarding the antibiotic activity of 6-aminoquinolones. The unique framework based on principal component analysis, projection onto latent structures, and response surface methodologies permits the structure-activity correlations to be shown and to suggest new compounds for further testing. The low activity of the suggested molecules points out the limitations of quantitative structure-activity relationship models when the training set is not properly designed in order to balance all the structural variations taken into account.