Gene expression profiling by cDNA array analysis in melanoma is hampered by the need for large amounts of RNA to prepare reliable probes for array hybridization. On the other hand, for ex vivo analysis of malignant cells from melanocytic tumors laser pressure catapulting is an essential prerequisite to obtain noncontaminated melanocytic preparations; however, laser pressure catapulting prepared material provides only nanogram amounts of RNA. In this study we present an approach to overcome these limitations by combining laser pressure catapulting and real-time polymerase chain reaction based SMART cDNA amplification technology. Reproducible and reliable hybridization patterns from about 500 laser pressure catapulting prepared cell equivalents from 22 cases of melanocytic tumors were generated using array analysis. Univariate analysis revealed significant differences of the expression pattern of melanocytic nevi, melanomas, and melanoma metastases. Multivariate analysis with four genes being the best univariate discriminative features (tyrosinase related protein 2, translation initiation factor 2 gamma, ubiquitine conjugating enzyme E2I and one expressed sequence tag) allowed clustering of nevi, melanomas, and melanoma metastases with an accuracy of 82%. Data validation was performed by additional quantitative reverse transcription-polymerase chain reaction (TaqMan-reverse transcription-polymerase chain reaction). Taken together, this study shows, that (1) array analysis is feasible on tumors with rather low cell numbers, and (2) differences in expression profiles allow discrimination between benign and malignant lesions. Expression patterns of marker genes defined in unequivocal histopathologic entities may improve the diagnostic and prognostic assessment of difficult melanocytic lesions, which is still the hardest problem in dermatopathology.