DNA Methylation Profiling Discriminates between Malignant Pleural Mesothelioma and Neoplastic or Reactive Histologic Mimics

Abstract

The diagnosis of malignant pleural mesothelioma (MPM) is challenging because of its potential overlap with other neoplasms or even with reactive conditions. DNA methylation analysis is effective in diagnosing tumors. In the present study, this approach was tested for use in MPM diagnosis. The DNA methylation patterns of a discovery cohort and an independent-validation cohort of MPMs were compared to those of 202 cases representing malignant and benign diagnostic mimics (angiosarcoma, desmoid-type fibromatosis, epithelioid sarcoma, leiomyosarcoma, lung adenocarcinoma, lung squamous cell carcinoma, melanoma, nodular fasciitis, reactive mesothelial hyperplasia, sclerosing fibrous pleuritis, solitary fibrous tumor, and synovial sarcoma). By both unsupervised hierarchical clustering and t-distributed stochastic neighbor embedding analysis, MPM samples in the discovery cohort exhibited a DNA methylation profile different from those of other neoplastic and reactive mimics. These results were confirmed in the independent validation cohort and by in silico analysis of the MPM-The Cancer Genome Atlas data set. Copy number variation profiles were also inferred to identify molecular hallmarks of MPM, including CDKN2A and NF2 deletions. Methylation profiling was effective in the diagnosis of MPM, although caution is advised in samples with low tumor cell content.

Feature Selection of OMIC Data by Ensemble Swarm Intelligence Based Approaches

“…The dataset GSE144910 collected a total of 88 genomic DNA samples taken from the postmortem superior temporal gyrus of the human brain with 44 schizophrenia and paired non-psychiatric controls ( Mckinney et al, 2020 ). The dataset GSE164269 generated 33 discovery and 46 independent validation cohorts of malignant pleural mesothelioma samples ( Bertero et al, 2021 ). The dataset GSE166787 contrasted DNA methylation data throughout human muscle cell differentiation in 28 individuals with type 2 diabetes and 28 controls ( Davegårdh et al, 2021 ).…”

Methylation classifiers: Brain tumors, sarcomas, and what's next

“…Many studies have been performed employing this platform. 3,5,8,9,[45][46][47][48][49][50][51][52][53][54][55][56][57][58][59] Several tissue sources, fresh, frozen, or FFPE embedded samples, are suitable for analysis.…”

“…DNA methylation profiling shows a clear separation of pleomorphic sarcomas from nonmesenchymal neoplasms such as mesothelioma and melanoma, with the latter exhibiting distinct methylation signatures. 54,86,87 On the other hand, several tumor types, including undifferentiated pleomorphic sarcomas, myxofibrosarcomas, and many pleomorphic liposarcomas, are almost indistinguishable by their DNA methylation patterns. 45 Furthermore, sarcomas may exhibit a methylation profile differing from that of the respective reference set resulting in a failed prediction.…”

“…The Illumina Infinium platform with its high‐density DNA methylation arrays, more specifically the HumanMethylation450 with 485 577 CpGs (released in 2011) and its successor, the MethylationEPIC with 853 307 CpGs (released in 2015) has been experienced as simple to use and generating reproducible data. Many studies have been performed employing this platform 3,5,8,9,45–59 . Several tissue sources, fresh, frozen, or FFPE embedded samples, are suitable for analysis.…”

“…The “sarcoma classifier” in its current version v12 contains reference sets for 62 tumor methylation classes. DNA methylation profiling shows a clear separation of pleomorphic sarcomas from nonmesenchymal neoplasms such as mesothelioma and melanoma, with the latter exhibiting distinct methylation signatures 54,86,87 . On the other hand, several tumor types, including undifferentiated pleomorphic sarcomas, myxofibrosarcomas, and many pleomorphic liposarcomas, are almost indistinguishable by their DNA methylation patterns 45 .…”