MicroRNAs (miRNAs) are a recently discovered class of small, non-coding RNAs involved in silencing of the expression of specific genes. Because miRNA expression is highly tissue specific, the miRNA profile of an unknown sample can be used to identify its origin. Around 3–5% of all newly diagnosed metastatic cancers are carcinoma of unknown primary (CUP) origin, where the site of the primary tumor cannot be determined, often because CUPs are poorly differentiated and hard to classify with current histological methods. By applying a novel microarray platform that is based on locked nucleic acid (LNA) modified detection probes, which enable very sensitive and specific detection of small RNA targets like miRNAs, we are able to classify tumors according to their tissue origin. We show that sample clustering in the miRNA space reflects the biological characteristics of the metastases, and could, potentially provide a basis for targeted therapy.