MicroRNA signature distinguishing nevi from primary melanoma

MicroRNA signatures have been previously used to detect melanoma progression from early to advanced metastatic stage. Here we sought to assess if differential expression of microRNAs could be used to distinguish benign nevi from primary melanomas. Archived skin lesions with paired adjacent melanoma and precursor nevus, having shared genotypes, were analyzed for differentially expressed miRNA by next generation sequencing. Using machine learning methods in combination with a random forest classifier, a signature of 14 miRNA was developed, that predicts melanoma with a 99% sensitivity and a 93% specificity. This signature was further validated on four independent, previously reported cohorts, which used microarray platforms, demonstrating the robustness of this signature across platforms. The utility of this signature as a non-invasive, pre-biopsy, screening tool for melanoma was explored. MicroRNA was collected from the skin surface of nevi and melanoma before biopsy, and a subset of 6 miRNA was found to classify melanomas with a performance AUC of 0.82, comparing favorably to the sensitivity and specificity of dermatologists visual inspection. These pilot data indicate that a miRNA based, non-invasive screen for melanoma is a potentially viable method for the objective detection of melanoma.