Background: The current AJCC TNM staging system has poor specificity for predicting visceral metastatic risk in patients diagnosed with stage I or stage II cutaneous melanoma. We, therefore, developed a gene expression profile signature (GEP) following in silico investigation of previously published microarray analyses. Methods: 60 formalin fixed paraffin embedded primary cutaneous melanoma samples from patients with stage I or II cutaneous melanoma with at least a follow up period of at least 6 years were macrodissected and analyzed blindly. RNA was isolated, converted to cDNA and RT-PCR was performed to assess the expression of the gene set. Expression data and biostatistical analysis was performed using GeNorm and JMP Genomics (SAS) Predictive modeling included Radial Basis Machine (RBM) and Partition Tree Analysis (PTA) Metastasis-free survival (MFS) was assessed using Kaplan-Meier analysis. The following clinical data was retrieved from medical records: survival, metastases, types of metastases. 20 out of 60 patients had developed visceral metastases in the follow up period. Results: GEP was developed following multiple analytical approaches.Two types of signatures emerged: Low risk (Class 1) and High risk (Class 2). Without optimizing for sensitivity, the analyses of the 60 sample cohort by radial basis machine (RBM) resulted in 92% ROC (met. accuracy = 90%, non-met. accuracy = 85%), while partition tree analysis (PTA) yielded 99% ROC (met. accuracy = 100%, non-met. accuracy = 95%). RBM classification showed 6-year MFS rates of 97% for Class 1 and 19% for predicted Class 2 of metastasis (median MFS = NR and 5.6 yrs, resp., P<0.0001 Log-Rank respectively). PTA showed 6-year MFS rates of 100% for predicted Class 1 and 14% for Class 2 of metastasis (median MFS = NR and 5.4 yrs, resp., P<0.0001 Log-Rank respectively). Conclusions: This study shows that DecisionDx-Melanoma GEP signature can provide excellent accuracy in predicting metastatic risk in stage I and II cutaneous melanoma.To our knowledge, the GEP provides the most accurate predictor to date for development of visceral metastases in patients with Stage I and II cutaneous melanoma.