Poster Presentation Australasian Society for Dermatology Research 2022 Annual Scientific Meeting

Investigating proteomic tools to find biomarkers in FFPE samples to differentiate between benign pigmented lesions and melanoma (#17)

Rachel Teh 1 2 3 , Ali Azimi 1 2 3 , Marina Ali 1 2 3 , Graham J Mann 4 , Pablo Fernández-Peñas 1 2 3
  1. Centre for Cancer Research, Westmead Institute of Medical Research, The University of Sydney, Westmead, NSW, Australia
  2. Westmead Clinical School123, Faculty of Medicine and Health, The University of Sydney, Westmead, NSW, Australia
  3. The Department of Dermatology, Westmead Hospital, The University of Sydney, Westmead , NSW, 2145, Australia
  4. The John Curtin School of Medicial Research, College of Health and Medicine, Australian National University, Canberra, ACT, Australia

Background: Benign melanocytic (pigmented) lesions can be difficult to distinguish from melanoma. Currently, biopsy and histopathology are the gold standard however uncertainty can lead to overdiagnosis or worse under diagnosis. There is a demand to improve diagnosis, understand melanoma, and its progression. Archival FFPE samples could be a valuable resource in proteomic investigations to identify biomarkers to profile melanocytic and melanoma lesions.

Aim & Objectives: To investigate archival FFPE samples and proteomic techniques and profile pigmented lesions to find biomarkers and diagnose melanoma

Method: The epidermis layer of the skin was extracted via LCM from archived de-identified FFPE tissues. Samples consisted of non-lesional (n=22), benign junctional naevi (n=22), dysplastic naevi (n=22), melanoma in situ (n=22), and minimally invasive melanomas <0.8mm (n=22). Following preparation and proteolytic digestion, the samples’ proteome was analysed in a TripleTOF mass spectrometer using SWATH-MS workflow. Differential abundance and machine learning (KNN, RF, SVM, DLDA) cross validation assessed classification potential for melanoma differentiation from benign lesions. Then, PCA and k-means clustering was used to visualise features, and improve classification. Finally, bioinformatic analysis using pathway analysis (IPA) and GO/Panther ontology identified enriched pathways and biological functions.

Results: We identified 4864 proteins across the 110 samples. Cross validation balanced accuracy of the most significant differentially abundant proteins (p<0.005, FC >1.5, <-1.5) revealed excellent performance (RF: ~88-93%) of the model. Further, k-means clustering revealed intermediate lesions (most heterogeneity) which closely correlated with either benign or malignant clusters. Bioinformatic analysis predicted altered proteins are involved in adhesion, cellular proliferation, apoptosis, invasion, and development of malignant tumours.

Conclusion:  This preliminary study demonstrates the successful detection of proteins in FFPE samples of benign naevi, dysplastic naevi, melanoma in situ, and early melanoma using SWATH-MS methodology. Exploration of this technique show a promising method for the detection of novel biomarkers in FFPE resources.