Multiple naevi are an important risk factor for melanoma, and naevi are precursor lesions for melanoma development. Despite recent advances in melanoma therapeutics, the only known cure is early detection and surgery for complete excision, prior to regional or distant spread of melanoma. Surveillance of existing naevi represents a significant opportunity for early diagnosis and treatment of high-risk lesions. This study comprises an early investigation as part of a larger study aimed at developing tools for sophisticated, molecular-based risk stratification of melanocytic naevi in the clinical setting.
Initially a literature search was performed to identify markers associated with malignant transformation. These markers were then tested across tissue microarrays (TMAs) to determine if they were differentially expressed in benign and malignant lesions, specifically at the dysplastic naevus-thin melanoma divide. Informative markers were then tested on a series of full width specimens. Statistical analysis was performed to determine which markers were useful when considered independently, and to determine a panel of markers that could distinguish between benign and malignant lesions.
Our results indicate that many of the included molecular markers, when considered in isolation are differentially expressed in benign and malignant naevi. Furthermore, a relatively small marker set is able to distinguish between diagnostic categories when trained and tested on the TMA dataset and the full width lesion dataset independently. A model trained on a TMA dataset and tested on the full width lesion dataset does not predict malignant diagnosis, but is able to predict which lesions are benign. This is likely due to the highly heterogenous nature of the lesions included in the full width lesion set.
This study has given us insight into the molecular basis of melanoma; specifically it has highlighted high mobility group protein B2 (HMGB2) as a potential driver of malignant transformation in melanoma. Overall, modeling was able to predict a benign lesion based on its molecular profile, but was very poor at selecting malignant lesions. Despite this, our findings will form the basis of a broader study aiming to develop a tool for molecular-based risk stratification that will aid early diagnosis and treatment of melanoma.