A Droplet Digital Polymerase Chain Reaction–Based Tool to Aid in Melanoma Diagnosis: Development of a 4-Gene Panel Using 164 Melanocytic Neoplasms Open Access

This study followed the standard criteria for the development of an analytical assay set forth by the College of American Pathologists. Institutional review board approval was obtained (Dartmouth Health, Department of Pathology and Laboratory Medicine, Lebanon, New Hampshire; IRB number 00031828).

We collected 113 FFPE skin samples submitted to the surgical pathology department at a large academic referral center from 2013 to 2022 that already had extracted DNA available because of CMA or next-generation sequencing performed for clinical care. The inclusion criterion was a diagnosis of melanocytic neoplasia made by at least 1 board-certified dermatopathologist. We defined borderline lesion as any melanocytic neoplasm not signed out as a benign nevus or a melanoma; most were dysplastic or atypical Spitz nevi (Table 1).

Cases that require CMA for clinical use are often borderline neoplasms. Thus, to avoid biasing the data set with borderline neoplasms, an additional 51 unambiguous melanomas and nevi were added. All 5 dermatopathologists selected the 51 additional melanocytic neoplasms, wherein the inclusion criterion was that CMA results would not change the final diagnosis, based on a review of the report and in the judgment of the dermatopathologist associated with the case. Thus, our total sample size comprised 164 (113 + 51) melanocytic neoplasms.

Prior to this study, a few limited experiments (n = 10) indicated that ddPCR copy number data may be most reliable in samples for which the time from embedding to DNA extraction is less than 1 month. This observation could likely be explained by the detrimental effects of formalin-induced chemical modifications and storing room temperature on the quality of extracted DNA.¹⁹  To clarify: in our experience, the skin sample itself can be several years old, provided that the DNA was extracted within 1 month of biopsy. We therefore restricted our analysis to such samples. DNA was isolated from tissue sections using the QIAGEN QIAamp FFPE Tissue-Kit (Hilden, Germany). Nucleic acid concentration was measured using a Qubit Fluorometer 3.0 and the Qubit dsDNA High-Sensitivity assay kit (ThermoFisher Scientific, Waltham, Massachusetts). All cases in our cohort had a DNA input of 10 ng.

CMA was performed using the OncoScan FFPE Assay Kit (Affymetrix, ThermoFisher Scientific). For samples with CMA-confirmed CDKN2A loss, the Chromosome Analysis Suite (ThermoFisher Scientific) software was used to determine whether losses were monoallelic or biallelic.²⁰  In keeping with our laboratory practices for CMA, all cases had at least 20% tumor cell content. All results were interpreted by an expert in the use of CMA for clinical purposes.

We targeted the CNVs of the melanoma FISH panel (RREB1 gain, CDKN2A loss, MYC gain, and MYB loss). FISH testing was not performed as part of this study; instead, we decided to use CMA, for several reasons. First, both CMA and ddPCR are run on extracted DNA, whereas FISH is not. Thus, any comparison of ddPCR with FISH would require cutting of additional unstained slides, which would introduce problems of tumor heterogeneity (due to different areas of inflammation, background nevi, etc). By using the same vial of homogeneous, extracted DNA for both CMA and ddPCR, such variation can be minimized. Second, we anticipate that future studies may use next-generation sequencing in follow-up research. If so, the extracted material would already be available. Finally, prior publications have revealed that CMA has sensitivities and specificities either superior or equivalent to those of melanoma FISH.^21,22 

Melanoma FISH assays sometimes include the cyclin D1 gene (CCND1) at 11q13. However, in prior published work, McFadden et al²³  found that gain of CCND1 may not be frequent, and so may not be as useful as previously believed. Thus, for this study, the final target loci were RREB1, MYC, MYB, and CDKN2A.

The study used a total of 164 skin samples. First, reference genes were selected, and 134 samples were used to develop the singleplex (ie, one target gene) assays (Figure 1, A and B). Then, the remaining 30 samples (16 for MYC and MYB; 14 for RREB1 and CDKN2A) were used to compare the performance of each gene’s singleplex assay with that of the multiplex assay (Figure 1, C and D). Finally, the multiplex panel was deployed on 65 skin samples—those with sufficient remaining DNA for ddPCR analysis—of the original 164 (Figure 1, E and F). At no point was any material re-extracted from the tissue block.

To obtain data on copy number, as opposed to absolute concentration, it is necessary to use a reference gene. The chief challenge of this process was to locate reference genes with minimal, or no, CNV as the melanoma evolves. This has already been done in a series of prior studies by O’Hern et al.¹⁶  To avoid loss or gain of one reference gene skewing the final target gene to reference gene ddPCR ratio, we used a panel of 7 different genes: threonine synthase-like 2 (THNSL2; assay ID dHsaCP2506297), lipase I (LIPI; dHsaCNS873986483), ribonuclease P RNA component H1 (RPPH1; assay ID dHsaCNS360987081), SLAIN motif family member 2 (SLAIN2; assay ID dHsaCP2506299), eukaryotic translation initiation factor 2C, 1 (EIF2C1; assay ID dHsaCP2500316), elongation factor Tu GTP binding domain containing 2 (EFTUD2; assay ID dHsaCNS493413543), and ribosomal protein lateral stalk subunit P0 (RPLP0; assay ID dHsaCNS163760503). To ensure reproducibility, each target gene to reference gene pair was run twice, often on different days by different individuals.

Assay IDs for the 4 target genes were RREB1, assay ID dHsaCP2506538;s CDKN2A, assay ID dHsaCP2506499; MYC, assay ID dHsaCNS260857412; and MYB, assay ID dHsaCP2506541. In keeping with the Minimum Information for Publication of Quantitative Real-Time PCR Experiments guidelines, the Minimum Information for Publication of Quantitative Real-Time PCR Experiments context sequence—a larger sequence that contains our smaller probe sequence—for all target and reference genes may be found in Supplemental Table 1 (see the supplemental digital content, containing 5 tables and 2 figures, at https://meridian.allenpress.com/aplm in the May 2025 table of contents). The detailed methods and results of these singleplex studies have been published separately.^16–18 

Although reliable data can be obtained from a series of such singleplex runs, that method has the drawback of occupying several wells per target gene. Fewer wells per sample can result in faster TATs because many samples can be run on one plate. To investigate whether multiple target and reference genes can be reliably assayed in a single well (“multiplexed”), 14 cases with CMA aberrancies in either RREB1, CDKN2A, or both were selected, and their copy number data from singleplexed versus multiplexed runs were compared. A similar exercise was performed for MYC and MYB genes, using an additional 16 cases.

Lastly, the droplet clusters in the corresponding 2D plot were assessed to ensure adequate droplet separation. When samples were multiplexed with the reference genes LIPI and SLAIN2, the concentration of one of the fluorescein amidite probes was diluted to 60% to shift the cluster away from the 100% concentration fluorescein amidite probe to ensure adequate separation.

ddPCR was performed using the Bio-Rad QX200 (Bio-Rad, Hercules, California). Per the manufacturer’s instructions, approximately 20 000 droplets were generated on the Bio-Rad QX200 droplet generator per reaction. Emulsified PCR reactions were run in a 96-well plate on the C1000 Touch Thermal Cycler (Bio-Rad), beginning with incubation at 95°C for 10 minutes, followed by 40 cycles of 94°C for 30 seconds, 60°C for 60 seconds, 10 minutes incubation at 98°C, and a final hold at 4°C for 1 hour. All PCR products were read on the Bio-Rad QX200 droplet reader, and copy number data were generated using the QuantaSoft software from Bio-Rad. The plates were analyzed on a Bio-Rad QX200 droplet reader with QuantaSoft v1.7.4 software (Bio-Rad) to determine the number of droplets positive for target and/or reference genes. Further details of the procedure have been published previously.^16–18 

To assess diagnostic performance, the total ddPCR panel was defined as “positive” if any one of the 4 genes assessed showed an aberrant copy number. For the purposes of this analysis, for simplicity, benign nevi and borderline lesions were grouped together under “nevi,” and both primary and metastatic melanomas were grouped together under “melanomas.”

Patient demographic, clinical, and histologic data were collected, including lesion size and location, TNM staging, Breslow depth (where applicable), and final diagnosis. One case (case 14) was a consult case wherein the slides were returned to the referring institution and were thus not available for the correlation with histologic features portion of our study. All other cases (n = 64) were examined for hemorrhage, necrosis, density of inflammatory infiltrate, presence of background nevus, nodular pattern of growth, sheetlike pattern of growth, and abundant cytoplasm as evaluated by a board-certified dermatopathologist. We defined abundant cytoplasm as cytoplasm larger than approximately 4 lymphocyte nuclei.

In the singleplexed ddPCR runs, the individual target gene to reference gene ratios were tabulated. Appropriate cutoff values were determined by receiver operating characteristic (ROC) curve analyses. These statistical analyses for MYC, MYB, RREB1, and CDKN2A targets were essentially identical. Portions of the singleplex data for RREB1, CDKN2A, MYC, and MYB have been published, and the statistics are described in more detail therein.^16–18 

In the multiplexed ddPCR runs, we calculated the ratio of target gene copy number (RREB1, CDKN2A, MYC, MYB) to reference gene copy number (THNSL2, LIPI, RPPH1, SLAIN2, EIF2C1, EFTUD2, RPLP0) for each sample. Then, for each target gene, the median of all ddPCR outputs (ie, the median of all target gene to reference gene ratios) was recorded. These median ddPCR outputs were used to perform the ROC analysis and to designate RREB1 gain, CDKN2A loss, MYC gain, and MYB loss and for all subsequent analyses.

For each target gene, the 65 FFPE samples were randomly split into 2 groups: a training cohort and a validation cohort. We used the training cohorts to determine the optimal median ddPCR cutoff values (Supplemental Figure 1, A through D); we then used the validation cohorts to evaluate the performance of these cutoffs on new, “unseen” samples (Figure 2, A through D). The cutoffs were determined and validated using standard ROC curve analysis methods. All statistical analyses were performed using R software (R Core Team, Vienna, Austria, 2013).²⁴ 

To investigate correlations between target gene copy number and degree of tumor progression, we calculated the median ddPCR copy number value for each gene in each diagnostic category (benign, borderline, primary melanoma, metastatic melanoma). Correlations between target copy number and degree of tumor progression were assessed for statistical significance using a Student 2-sample t test for means.

To investigate correlations between histologic features and target gene CNVs, each target gene’s positive or negative CNV status—determined by CMA—was tabulated and correlated against the aforementioned histologic features. To assess the statistical significance of these correlations, a Fisher exact test was performed on each gene–histologic feature pair. All statistical tests were 2-sided, and statistical significance was considered P < .05.

Results of the singleplex data on the target genes have been published separately^16–18  and show high concordance with CMA.

Comparison of singleplex versus multiplex data for RREB1, CDKN2A, MYC, and MYB revealed similarly high concordance: for each sample, the median copy number value obtained from its multiplex ddPCR run was within ±0.2 of its corresponding singleplex ddPCR run.

For the multiplexed ddPCR runs, 65 FFPE samples were used, comprising 3 benign nevi (4.6%), 11 borderline lesions (16.9%), 33 primary melanomas (50.8%), and 18 metastatic melanomas (27.7%). Of these 65 FFPE samples, 14 (21.5%) had CMA performed as part of clinical care. For the remaining 51 cases (78.5%), CMA was performed for research purposes only, and was not involved in establishing the diagnosis. Demographic, clinical, histologic, and diagnostic information is summarized in Table 1 and Supplemental Tables 2 and 3.

For the training cohorts, the sample sizes were n = 32 (RREB1), n = 31 (CDKN2A), n = 32 (MYC), and n = 31 (MYB). The numbers of CMA positive/CMA negative cases (ie, CNV detected/CNV not detected) were 14/18 for RREB1 gain, 14/17 for CDKN2A loss, 9/23 for MYC gain, and 6/25 for MYB loss. ROC analyses revealed optimal median ddPCR cutoffs of 2.36, 1.63, 2.67, and 1.17 for RREB1 gain, CDKN2A loss, MYC gain, and MYB loss, respectively (Supplemental Figure 1, A through D).

For the validation cohorts, the sensitivity and specificity data are summarized in Table 2. Expanded ddPCR and CMA data for each case can be found in Figures 3, 4, 5, and 6, as well as in Supplemental Table 4.

Given the importance of distinguishing between monoallelic and biallelic loss of CDKN2A,⁵  a second ROC analysis for CDKN2A was performed. This revealed an optimal biallelic CDKN2A loss cutoff of 1.05 (median ddPCR value <1.05 = biallelic CDKN2A loss). Using this cutoff, ddPCR correctly identified all 12 cases of biallelic loss (Figure 5). In all 4 atypical Spitz tumors (ASTs), the monoallelic status of CDKN2A was accurately detected by ddPCR.

In our series, the sensitivity and specificity of CMA for distinguishing nevi from melanoma were 86.2% and 78.6%, respectively. ddPCR showed comparable sensitivity (78.4%) and specificity (71.4%) for this purpose. These sensitivities and specificities should not be confused with those in Table 2. The former are diagnostic in nature (ie, whether ddPCR results, when used in the same manner as melanoma FISH panels, correctly distinguish nevi from melanomas); the latter—those listed in Table 2—are analytic in nature (ie, whether ddPCR and CMA results agreed on the CNV status of one gene).

By individual gene, ddPCR was highly concordant with CMA: concordance rates for RREB1 gain, CDKN2A loss, MYC gain, and MYB loss were 92.3% (60 of 65 cases), 92.3% (60 of 65 cases), 93.8% (61 of 65 cases), and 95.3% (62 of 65 cases), respectively. The concordance rate between the overall 4-gene ddPCR panel and the CMA data was 92.3% (60 of 65 cases).

Our data revealed a positive correlation between ddPCR median RREB1 values and the degree of tumor progression (Figure 7). Benign nevi had essentially no REBB1 gain (mean ddPCR value, 1.95; SD = 0.17), whereas borderline lesions exhibited some gains (mean ddPCR value, 1.96; SD = 0.24). Gains became more dramatic in melanomas (mean ddPCR value, 2.45; SD = 0.49), and even more so in metastases (mean ddPCR value, 2.68; SD = 0.62). Each pair is depicted in Figure 7; the ddPCR values for MYC, MYB, and CDKN2A did not reveal similar correlations with tumor progression in our data.

Histologic evaluation revealed that 11 of the 18 cases with MYC gain by CMA (61.1%) and 7 of the 47 cases with no MYC gain by CMA (14.9%) had a sheetlike pattern of growth. This difference was statistically significant (odds ratio [OR], 6.98; P = .002). Similarly, 8 of the 14 cases with MYB loss by CMA (57.1%) had a sheetlike pattern of growth, compared with 11 of the 51 cases with no MYB loss by CMA (21.6%). This difference was also statistically significant (OR, 4.47; P = .02) (Supplemental Table 5).

Seventeen of the 28 cases with RREB1 gain by CMA (60.7%) and 12 of the 37 cases with no RREB1 gain by CMA (32.4%) exhibited expansile, nodular growth. This difference was statistically significant (OR, 2.91; P = .045). Similarly, 14 of the 18 cases with MYC gain by CMA (77.8%) showed a nodular pattern of growth, whereas 15 of the 47 cases with no MYC gain by CMA (31.9%) had that pattern. This difference was also statistically significant (OR, 6.77; P = .002) (Supplemental Table 5).

Eleven of the 18 cases with MYC gain by CMA (61.1%) and 9 of the 47 cases with no MYC gain by CMA (19.1%) had epithelioid cells with an abundant cytoplasm. This difference was statistically significant (OR, 6.06; P = .003). Similarly, 8 of the 14 cases with MYB loss by CMA (57.1%) had this histologic feature, compared with 12 of the 51 cases with no MYB loss by CMA (23.5%). This difference was also statistically significant (OR, 4.01; P = .047) (Supplemental Table 5).

Histopathologic features for all 64 cases (excluding case 14, the one consult case as indicated in cohort selection), along with the complete statistical analysis, are included in Supplemental Tables 3 and 5, respectively.

Diagnosing melanocytic neoplasms that have ambiguous histopathologic findings can be challenging. In this study, we show that diagnostically useful copy number data may be obtained using a newer, faster molecular method, ddPCR. We now discuss the performance and potential use of a 4-gene ddPCR panel, hypothesize regarding the morphologic-molecular correlations, and highlight directions for future research.

Our data revealed that both sensitivity and specificity of the ddPCR 4-gene panel were within approximately 7% of those of CMA in the diagnosis of melanoma. The performance of ddPCR in detecting CNV was similar to that of melanoma FISH in the literature: that is, the concordance between ddPCR and CMA in our study was 92%, and the published concordance rate between melanoma FISH and CMA is 91%.²⁵ 

There were 5 cases (7.7%) in which ddPCR was discrepant with CMA: 4 false negatives and 1 false positive. A systematic investigation of all the possible causes of these results is beyond the scope of the study. However, 3 of the 4 false negatives had associated necrosis, which has been documented to negatively impact the purity of isolated DNA from FFPE tissue.²⁶  Further study is warranted to indicate whether ddPCR false negatives can be minimized by selecting sections without necrosis. Notably, all 4 false-negative cases were included as research cases. In practice, such necrotic and “histologically obvious” melanomas would likely not require a ddPCR panel for diagnosis.

Notably, the sensitivities of ddPCR in detecting CDKN2A loss (80.0%), MYC gain (70.0%), and MYB loss (71.4%) were lower than their corresponding specificities (all 100%). In other words, ddPCR may be more susceptible to false negatives than to false positives. In clinical practice, users should be aware that the sensitivities of MYC gain and MYB loss are lower than those of RREB1 gain and CDKN2A loss. However, in our cohort, 0 of the 65 cases had isolated MYC gain or isolated MYB loss. Thus, although theoretically possible, in practice, isolated MYC gain or isolated MYB loss leading to a false-negative 4-gene ddPCR panel result may not be common.

This underscores the importance of a comprehensive approach—considering histopathology, clinical evaluation, and immunohistochemical staining alongside molecular studies—when faced with diagnostically challenging lesions. This is especially critical in cases where ddPCR results are negative, yet clinical suspicion remains high. We emphasize that ddPCR, if eventually used in clinical practice, would be part of a multimodal diagnostic strategy, not a stand-alone test.

Concerning cellularity and tumor purity, the technique was successfully deployed on melanomas with a broad range of Breslow depths, from 0 mm (melanoma in situ) to 16 mm. All cases had at least 20% tumor cell content, only because that is the requirement for CMA in our laboratory.

With further validation and optimization, a multiplexed 4-gene ddPCR panel could become an additional diagnostic tool for melanoma. The choice of whether to use CMA, FISH, or ddPCR could be decided on a case-by-case basis such that the most relevant information is obtained. We emphasize that ddPCR, if eventually integrated into clinical practice, would be used in addition to (not in replacement of) CMA and FISH.

The increases in RREB1 copy number with respect to degree of tumor growth, although consistent with known tumor biology, have not been previously reported to our knowledge. RREB1 gain is the most common CNV in melanoma.¹⁶  In that context, our results would be explained by a model wherein RREB1 gain occurs early, persists via natural selection, and exacerbates as the tumor progresses. Proving such a hypothesis requires future inquiry beyond the scope of this study. Future work may also reveal whether median ddPCR RREB1 copy number can be used to predict prognosis or therapeutic response.

Studies have suggested that CDKN2A deletion is most clinically significant when biallelic.^27,28  In a seminal paper, Gerami et al²⁸  used a 4-probe FISH panel on atypical Spitzoid neoplasms and found a positive correlation between biallelic CDKN2A deletions and distant metastases. In our cohort, ddPCR successfully identified almost all cases of monoallelic loss of CDKN2A, including all 4 AST cases. The number of ASTs was too few to draw definite conclusions, but the ddPCR assay’s performance suggests that it could be of value, and may be worth investigating, for predicting the metastatic potential of ASTs.

It is also noteworthy that the 1.05 optimal median ddPCR cutoff for biallelic CDKN2A loss is almost exactly what would be expected from the tumor biology. Theoretically, in a sample composed only of tumor cells with monoallelic CDKN2A loss, the CDKN2A copy number per cell should be 1.00. Because of the heterogeneous nature of any tumor, a shift away from that cutoff value of 1.00 is expected. From our data, the shift is present, but minimal, amounting to 0.05 target gene copies. Although ddPCR has been used to distinguish between monoallelic and biallelic CDKN2A loss in other tumors,²⁹  this work is the first to make such a distinction in melanoma.

These data add to the limited literature on the correlation between MYC and morphology in melanoma. In a FISH study of 40 melanomas, all of which had gain of MYC, a nodular pattern of growth was found to be frequent.^30,31  Our data corroborate and extend that observation; when directly compared with melanomas that lack gain of MYC, our study finds a nodular pattern of growth to be a statistically significant discriminator. In that prior literature it was suggested that, because MYC is a powerful stimulator of cellular proliferation,³¹  gain of the gene may result in proliferative foci, which evolve into a nodular pattern of growth. Our data would be in keeping with that model. Moreover, ddPCR also reveals that MYC copy number continues to escalate within metastases, possibly conferring an enduring survival advantage to that cell population. A representative histologic image of a melanoma with MYC gain, along with its ddPCR and CMA data, is shown in Figure 8, A through C.

In other tumors, MYB usually acts as a proto-oncogene. For reasons that remain poorly understood, deletion of this gene can promote tumorigenesis in melanoma and a small number of other cancers.³²  We found that deletion of MYB had a statistically significant association with a sheetlike pattern of growth. To our knowledge, this association has not been previously reported.

The reason for this association is not immediately clear, but could be correlation rather than causation. That is, all the melanomas that had deletion of MYB in our cohort were large, cellular tumors. Almost half (7 of 15) were metastases. It is possible that both loss of MYB and a sheetlike pattern of growth are late events that occur in advanced melanomas. To delineate the precise relationship, further study is necessary. A representative histologic image of a melanoma with MYB loss, along with its ddPCR and CMA data, is shown in Supplemental Figure 2, A through C.

Strengths of this study include the broad variety of clinical samples, comprising a wide range of melanocytic neoplasms obtained from both pediatric and adult patients. Additionally, the BioRad ddPCR system used in this study is widely available in hospital laboratories and is used in various research and clinical applications, such as virus detection,³³  germline copy number carrier screening assays,³⁴  and liquid biopsy cancer mutation detection.³⁵  For ddPCR systems besides the one we used, we anticipate that the reference genes identified in our study should still suffice. Also, newer commercially available ddPCR platforms allow for increased multiplexing. These could also be adopted for use with the probe sets used in the current study. These have the potential for higher reagent efficiency, leading to even lower costs per patient.

One limitation of our study is that it was conducted at a single institution. We cannot rule out the possibility that subtle differences in fixation practices at other laboratories could impact the yield of DNA, and thus results; therefore, it is important that the work be reproduced at other institutions. Also, our sample size was restricted by factors such as (1) cost and time required for CMA analysis (this is why relatively few confidently benign nevi were included) and (2) measures taken to optimize the quality and quantity of isolated DNA (eg, DNA extraction was performed within 1 month of biopsy). Future research, with larger cohorts across a wide range of organizations, could investigate whether these inclusion criteria could be broadened. Finally, we abided by the 20% tumor cellularity cutoff for this study because, in our clinical laboratory’s experience, it is uncommon to have cases below this threshold; future studies could examine whether this limit of detection (LOD) could be lowered for ddPCR.

Molecular analysis of cancers based on acquired genomic aberrations is being implemented in both research and clinical care. Certain fields have seen greater advances than others: applications have expanded from aid in diagnosis to development of targeted therapy. However, high costs and long TATs remain challenging. This 4-gene ddPCR panel can provide data comparable to the gold standard, CMA. Future studies could specifically design experiments to include FISH testing alongside CMA and ddPCR. This would allow for a more comprehensive understanding of the discrepancies observed between ddPCR and CMA in our study: FISH could serve as a tiebreaker in such cases. In our work, all of which was performed in a clinical laboratory, data were available within 24 to 48 hours of ordering.

Potential applications of this 4-gene ddPCR panel are not limited to diagnostic pathology; indeed, it may also contribute to our understanding melanomagenesis and prognostication. Significantly, the quantitative nature of ddPCR results may be particularly useful in the study of tumor evolution. For instance, future studies could investigate whether specific copy number values for the target genes are associated with clinical stage and other important end points.

Future applications of this 4-gene ddPCR panel could also examine partially transformed neoplasms, objectively address the question of “how far along” the transformation is, and correlate ddPCR value with end points of interest (eg, probability of metastasis).

Moreover, the paradigm used in this study could serve as a template to design and implement ddPCR testing in other tumors. In our laboratory, we have begun work on applying the technique to help diagnose other diseases, including mycosis fungoides and differentiated vulvar intraepithelial neoplasia. CNV has been recently documented to be important in some pulmonary adenocarcinomas, in certain gliomas, and in chemotherapeutic resistance in colon cancer.^36–38  We believe other colleagues may wish to deploy the ddPCR technique on these and even more diseases.

The 4-gene multiplexed ddPCR panel, designed to detect CNV of RREB1, CDKN2A, MYC, and MYB to aid in the diagnosis of melanoma, has high sensitivity and specificity, comparable to those of CMA. We also report some noteworthy correlations with morphology. A fully validated form of this quantitative panel could be added to the molecular tool kit for the diagnosis and study of melanoma.

The authors would like to thank Mohammed Azim, PhD, for his insights and reviews of the manuscript, along with the members of the Pathology Shared Resource Laboratory, a section of the laboratory for Clinical Genomics and Advanced Technology (CGAT). The data in this study were in part generated through CGAT in the Department of Pathology and Laboratory Medicine of the Geisel School of Medicine at Dartmouth, the Dartmouth-Hitchcock Medical Center, and the Norris Cotton Cancer Center (NCI Cancer Support Grant #5P30CA023108-37).