Purpose: Pathologic diagnosis of melanocytic tumors can be difficult and prone to error. More accurate ancillary tools are needed. We present a genomic model for distinguishing benign (nevi) from malignant (melanoma) melanocytic skin tumors. Experimental Design: A DNA-based gene panel was evaluated on 250 nevi and melanomas, split into discovery and validation sets. Mutational loads were calculated, and a logistic regression model was developed and validated. An independent cohort of 110 borderline melanocytic tumors was analyzed against pathology-based tiers of malignancy and likely pathways of origin. Results: Combining non-BRAF/non-NRAS coding mutational load and noncoding mutational load, we obtained an AUC of 95% in the discovery and 96% in the validation cohort. Using a Youden index threshold, the resulting classifier demonstrated >95% specificity and >80% sensitivity in the discovery and validation cohorts. In the cohort of 110 borderline melanocytic tumors, we identified key driver mutations in agreement with the likely pathway of origin. Noncoding mutational load (P = 0.02) and melanoma probability (P = 0.02) from the genomic model were significantly associated with malignancy estimates in borderline tumors lacking pathway-defining genomic aberrations (conventional tumors). The genomic model estimated a high probability of melanoma in 8 of 36 (22%) conventional tumors with indeterminate malignancy, with 7 of 8 showing negative/nonaberrant results on other ancillary diagnostic tests. No malignant behavior based on the genomic model was seen in four conventional borderline tumors with >1 year of follow-up. Conclusions: This genomic model can become a clinically useful ancillary tool that is highly specific for differentiating melanomas from nevi.
29 March 2022 in Journal of the European Academy of Dermatology and Venereology
Distribution of total, coding and non-coding mutational load against pathological and clinical features. (A) Distribution in melanomas. (B) Distribution in nevi. LOESS regression is shown in blue for thickness, TMR and age at biopsy. NS: Not significant. (*): p-value<0.05. (**): p-value<0.01. (***): p-value<0.001.