Microbial resistance is an increasing health concern and a true danger to human wellbeing. A worldwide search for new compounds is ongoing and antimicrobial peptides are promising lead candidates for tomorrow's antibiotics. The decapeptide anoplin, GLLKRIKTLL-NH2, is an especially interesting candidate because of its small size as well as its antimicrobial and nonhemolytic properties. Optimization of the properties of an antimicrobial peptide such as anoplin requires multidimensional searching in a complex chemical space. Typically such optimization is performed by labor-intensive and costly trial and error. In this study we show the benefit of fractional factorial design for identification of the optimal antimicrobial peptide in a combination matrix. We synthesize and analyze a training set of 12 anoplin analogs, representative of 64 analogs in total. Using MIC, hemolysis and HPLC retention time data, we construct analysis of variance models that describe the relationship between these properties and structural characteristics of the analogs. We show that the mathematical models derived from the training set data can be used to predict the properties of other analogs in the chemical space. Hence, this method provides efficient identification of the optimal peptide in the searched chemical space.