OA08.07 In Silico Discovery of Unannotated miRNAs in Malignant Pleural Mesothelioma Reveals Novel Tissue-of-Origin Markers

Malignant pleural mesothelioma (MPM) is an aggressive disease. One of the major clinical challenges associated with MPM is the lack of biomarkers capable of distinguishing primary MPM from cancers that have metastasized to the pleura. The current gold standard consists of a panel of positive and negative protein markers to confirm tissue-of-origin; however, many cases remain undistinguishable from other thoracic cancers. Recent studies have suggested that the human genome encodes more microRNAs (miRNAs) than currently annotated. These undescribed sequences have been shown to display enhanced tissue and lineage specificity. Therefore, we hypothesize that MPM tumors express a specific set of previously unannotated miRNA sequences with tissue-specific expression capable of distinguishing MPM from other thoracic diseases. Novel miRNA candidates were detected from small RNA-sequencing data generated by The Cancer Genome Atlas (TCGA) (n=87 MPM) using the miRDeep2 algorithm, a well-established novel-miRNA prediction algorithm. The possible biological roles of these miRNA candidates were investigated by performing a genome-wide 3’UTR target prediction analysis. Additionally, their tissue-specificity was assessed using expression profiles of 1,093 lung tumors from four independent cohorts of adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC). Finally, we developed a miRNA-based classifier model using the weighted voting class prediction method to distinguish MPM from other thoracic cancers. Our initial analysis revealed 424 miRNA candidates, which were subsequently filtered by RNA structure, abundance of sequencing reads, and genomic location, resulting in 154 previously unannotated miRNA sequences. Interestingly, the novel miRNAs were predicted to target protein-coding genes involved in MPM biology, including the Ataxia Telangiectasia Mutated (ATM) gene, a tumour-supressor gene frequently mutated in MPM. Likewise BRCA1 Associated Protein 1 (BAP1), involved in the DNA damage response pathway, was also a predicted target. Principal component analyses revealed that novel-miRNA expression was able to distinguish MPM from LUAD and LUSC. Furthermore, our miRNA-based classifier model revealed 10 novel miRNAs capable of successfully identifying 86 out of the 87 MPM cases (98.80%) and 100% of LUAD cases (true positive rate = 98.85%, false positive rate = 1.150%). Here, we provide evidence for the presence of 154 previously unannotated miRNA species relevant to MPM. These miRNAs not only significantly expand the miRNA repertoire but also unveil specific roles in MPM biology. Most importantly, the strikingly high sensitivity and specificity of the novel miRNA-based classifier in distinguishing MPM from LUAD illustrates the potential of using these novel miRNAs to supplement current clinical markers to define MPM.