MSACL 2016 US Abstract

Molecular Diagnosis of Benign and Malignant Melanocytic Lesions Using Mass Spectrometry Imaging

Erin H Seeley (Presenter)
Protea Biosciences, Inc.

Bio: Erin Seeley received her PhD in Analytical Chemistry from Purdue University in 2005 under the direction of Fred Regnier. Her research focused on affinity selection of phosphopeptides and characterization by LC-MS/MS approaches. She then moved to Vanderbilt University where she was a postdoc in the newly founded Mass Spectrometry Tissue Imaging Core under the direction of Richard Caprioli. In 2006, Erin was promoted to Associate Director of the Tissue Imaging Core, a position she held until the end of 2013. During her time at Vanderbilt, Erin led projects with clinicians and clinical researchers on difficult diagnostic and prognostic problems resulting in over 35 peer reviewed publications. In 2014, Erin moved to Protea Biosciences to lead efforts in MALDI Mass Spectrometry Imaging and particularly in Clinical Mass Spectrometry Applications.

Authorship: Erin H. Seeley (1), Ralitza Gueorguieva (2), and Rossitza Lazova (2)
(1) Protea Biosciences, Inc., Morgantown, WV (2) Yale University School of Medicine, New Yaven, CT

Short Abstract

Mass spectrometry imaging (MSI) is an emergent technology for the analysis of clinical samples. Here, we present an application of MSI to the molecular diagnosis of human melanocytic skin lesions. A training set of 25 benign nevi and 25 malignant melanomas were interrogated and used to create a genetic algorithm for classification. This algorithm was then validated in an independent set of lesions (sensitivity 97%, specificity 85%). Ongoing work is looking at the application of the algorithm to melanocytic lesions that fall in the histologic “gray area” as well as identification of the peptides that were part of the classifier.

Long Abstract


The definitive diagnosis of malignant melanoma can be challenging in a number of cases and discordance rates between dermatopathologists can be quite high (1). In a previous study, we showed that mass spectrometry imaging (MSI) could accurately differentiate between one type of benign nevi called Spitz nevi (SN) and Spitzoid malignant melanomas (SMM) (2). Here we apply MSI technology to allow for a molecular differentiation between benign melanocytic nevi and conventional malignant melanomas. We sought to determine whether a proteomic molecular diagnostic approach could help improve the diagnosis and decrease the subjectivity of histopathologic examination, especially in cases that are not clear-cut.


A histology guided mass spectrometry imaging (HGMS) approach was applied to a cohort of archival formalin-fixed, paraffin embedded tissue samples consisting of 51 benign melanocytic nevi and 55 malignant melanomas. Only cases with a firm diagnosis of nevus or melanoma were included in the study. In this approach, two sections of each tissue sample were collected. One was placed onto a mass spectrometry target. A consecutive serial section was positioned onto a glass slide for hematoxylin and eosin staining and histopathologic review of digital microscopy images of the lesion. Areas corresponding to tumor (melanocytes) were annotated (~40 per sample; ~20 in the superficial portion of the lesion and ~20 in the deep portion) using color-coded circles of 100 µm diameter. Annotated images were merged with digital images of the unstained sections using Photoshop. The mass spectrometry sections were deparaffinized and antigen retrieved before being subjected to on-tissue tryptic digestion and matrix application. Mass spectra were collected from areas specified by the pathologist using a Bruker ultrafleXtreme MALDI TOF/TOF mass spectrometer.


A genetic algorithm classification model was created from a subset of the sample – training set, including 25 benign and 25 malignant lesions. Mass spectra were preprocessed including baseline subtraction, normalization, and alignment to common peaks using ClinProTools (Bruker). Peaks boundaries were manually determined and the peaks integrated. Approximately 850 monoisotopic peaks were detected from the dataset. Of these, 109 were found to be significant between the groups using a Bonferroni corrected p-value of 0.00006. Additionally, 181 peaks had areas under receiver operator curves of greater than 0.8. The optimized algorithm consisted of 24 peptide peaks and resulted in a sensitivity and specificity of 100% using a leave 20% out cross validation within the training set. The optimized algorithm was then applied in a validation mode to the remaining 26 benign nevi and 29 malignant melanomas. Within these samples, a sensitivity of 97% and a specificity of 85% were achieved.

The real test of the molecular classification algorithm is how well it can perform in cases where there is ambiguity in the histopathologic analysis. To this end, we are working on applying the classification algorithm to a cohort of these cases where clinical outcome is used as the gold standard. Cases are restricted to those with a minimum of 5 years of clinical follow up to determine if any adverse events occurred such as recurrence, lymph node and/or distant metastases, or death. These results are also being compared to results from fluorescence in situ hybridization and comparative genomic hybridization, when available.

Work is also underway to identify the proteins from which the peptides originate that are either statistically significant and/or part of the genetic algorithm classifier. Once identified, the protein expression levels can be validated using traditional immunohistochemistry approaches.


Mass spectrometry has been shown to accurately diagnose melanocytic lesions based on proteomic differences. This technology is likely to be helpful in cases where there is high discordance between pathologists, the so called “indeterminate” or “gray area” cases. Identification of these proteins that differentiate benign from malignant melanocytic lesions may lead to biomarker discovery that can aid in the diagnosis and treatment of patients.

References & Acknowledgements:

(1) Farmer ER, Gonin R, Hanna MP. Human Pathol. 1996, 27, 528-531.

(2) Lazova R, Seeley EH, Keenan M, Gueorguieva R, Caprioli RM. Am J Dermatopathol. 2012, 34, 82-90.

Financial Disclosure

SalaryyesProtea Biosciences
Board Memberno
ExpensesyesProtea Biosciences

IP Royalty: no

Planning to mention or discuss specific products or technology of the company(ies) listed above: