Treatment protocols for malignant tumors in the oral cavity differ greatly based on the presence of cervical lymph node metastasis. We applied gene expression profiling to the pathological lymph node status and used a training-test approach to evaluate the reliability of cDNA microarray-based classifications of 15 matched resected primary oral squamous cell carcinomas (OSCCs) and corresponding normal oral tissues. The clustering of all the microarray data was separated into two groups based on metastatic node positivity and node negativity. Furthermore, a 20-gene signature was identified that differentiated the testing set (n=8) with high classification accuracy (88%). Our findings support the hypothesis that the lymph node metastasis status can be predicted using the gene expression patterns of the primary OSCC, and may be a powerful tool in identifying patients at high risk of lymph node metastasis.