Discrimination of Benign Versus Malignant Melanocytic Skin Neoplasms by MALDI Imaging Mass Spectrometry (MALDI IMS)
Alireza Sepehr, Erin Seeley, Anna Harris, Steven Tahan, Richard Caprioli. Beth Israel Deaconess Medical Center & Harvard Medical School, Boston, MA; Vanderbilt University, Nashville, TN
Background: Diagnosis of melanoma is heavily based on histopathologic criteria and remains challenging in a large group of cases, as the significance of the criteria differ among dermatopathologists. Imaging mass spectrometry (IMS) combines the measurement capability of mass spectrometers with a surface sampling process that allows probing and mapping of the protein content of a sample. Using MALDI, we have applied IMS for the discrimination of benign versus malignant melanocytic skin tumors.
Design: Hematoxylin and eosin (H&E) slides and formalin fixed, paraffin embedded (FFPE) tissue blocks of invasive malignant melanomas (MM) and dermal nevi (DN) of patients with clinical follow-up data were retrieved from the archives of the BIDMC. One unstained section per case was prepared. Digital microscope image of the H&E sections were annotated and superimposed to an image of the unstained section. Trypsin and matrix were spotted onto the unstained sections at the annotated locations. Custom geometry files were created to allow for the targeting of the specific areas where trypsin and matrix have been applied and mass spectral profiles were collected directly from the tissue sections using MALDI TOF MS. A class prediction model was created using a support vector machine algorithm in ClinProTools software.
Results: 75 tumors, including 34 MMs and 41 DNs were selected. We generated IMS data and successfully profiled tissues using robotic matrix deposition with a 200 µm resolution and identified candidate peptide peaks (m/z range: 700-4500) which were preferentially over-expressed in cases of MM or in DN (p < 0.05). When we used a leave 20% out cross validation genetic algorithm classification with selected peptide peaks, we were able to reach an overall spectral classification of 95% in the training set (n=40). We then applied the candidate peptide peaks to a blindly-selected validation set (n=35) of MM and DN and reached 80% and 85% spectral classification accuracy, respectively.
Conclusions: We compared the peptide expression profile of MM and DN by MALDI IMS on FFPE tissues in separate training and validation sets. We developed algorithms with overall spectral accuracy of 84% for MM and 89% for DN classes. These spectral rates provide excellent case discrimination capabilities for the diagnosis of melanocytic tumors. IMS can be used as an adjunct to routine histopathologic and immunophenotypic evaluation of cutaneous melanocytic tumors.
Monday, March 4, 2013 11:00 AM
Proffered Papers: Section F, Monday Morning