Identification of novel RNA biomarkers for the differential diagnosis of cutaneous melanoma and nevi.

Abstract

9580 Background: The differential diagnosis of nevi and melanomas remains a significant clinical challenge, as misclassification can result in overtreatment or delayed care. Despite advances in clinical, dermatoscopic, and histopathological methods, benign lesions are frequently misclassified as nevi, SAMPUS, or MELTUMP, while atypical melanomas are often mistaken for nevi. This study aimed to identify novel RNA markers for the differential diagnosis of nevi and melanomas. Methods: Ninety histological samples of melanocytic neoplasms were analyzed, including 45 morphologically confirmed nevi and 45 melanomas. Massive parallel sequencing was performed using the NextSeq 550 system (Illumina, USA) with the “NextSeq High Output 150 Cycles Kit” reagent set, following the manufacturer’s protocol. Data normalization employed FPKM and TPM metrics, and differential gene expression analysis utilized 26 algorithms on the RNA-Seq 2G web server. Selected RNA markers were further validated on an independent cohort of 120 samples (60 verified nevi and 60 melanomas) using RT-PCR. Results: The study analyzed the expression of over 18,000 coding and 42,000 non-coding RNAs (primarily long non-coding) in histologically confirmed melanomas and nevi, balanced between "classical" and dysplastic nevi. Initial RNA-Seq quality control confirmed high data integrity and sufficient sequencing depth in 87 of 90 samples. Melanoma-specific markers identified included CSAG1, CXCL1, CXCL8, CXCL9, DUXAP8 + DUXAP9 + DUXAP10, FCRL3, IGHA1, IGHG1, LRP2, MAGEA3 + MAGEA6 + MAGEA12, MMP1, OR2I1P, SPP1, and VGF, while nevus-specific markers included CD44-AS1, CDR1-AS (LINC00632), DSCAS, and ENSG00000287270. For RNA marker testing at the next stage, all 120 samples were deemed suitable for analysis. Modeling of multimarker tests achieved an AUROC of 0.94, with a model incorporating only 5 preselected markers outperforming those with a larger number of markers. The most informative logistic regression model included the following marker combinations: MAGEA3 + MAGEA6, CXCL8 + LINC00632, DUXAP8 + DUXAP9 + DUXAP10, CSAG1, and CXCL1. Conclusions: The model, incorporating only 5 markers, achieved an AUROC of 0.94 in an independent validation cohort. It is now positioned for further validation in cohorts enriched with samples of uncertain malignant potential based on histological evaluation. Clinical trial information: NCT04353050 .