New Publication from Dr. Jüri Reimand
The paper describes an integrative analysis of prognostic non-coding RNAs (lncRNAs) in cancer using machine learning to identify lncRNAs that were indicative of poor patient survival in several types of cancer, and experimentally validated one novel lncRNA HOXA10-AS that functions as a potent oncogene in brain cancer. The study finds find novel leads for biomarker development and also finds that some lncRNAs may regulate cancer pathways and serve as potential targets for therapies.
Karina Isaev, a former MBP graduate student in Dr. Reimand’s lab, served as first author on the paper. The project was also a close collaboration with Dr. Daniel Schramek’s lab at the Lunenfeld-Tanenbaum Research Institute and the Molecular Genetics department.
Long non-coding RNAs (lncRNAs) are increasingly recognized as functional units in cancer and powerful biomarkers; however, most remain uncharacterized. Here, we analyze 5,592 prognostic lncRNAs in 9,446 cancers of 30 types using machine learning. We identify 166 lncRNAs whose expression correlates with survival and improves the accuracy of common clinical variables, molecular features, and cancer subtypes. Prognostic lncRNAs are often characterized by switch-like expression patterns. In low-grade gliomas, HOXA10-AS activation is a robust marker of poor prognosis that complements IDH1/2 mutations, as validated in another retrospective cohort, and correlates with developmental pathways in tumor transcriptomes. Loss- and gain-of-function studies in patient-derived glioma cells, organoids, and xenograft models identify HOXA10-AS as a potent onco-lncRNA that regulates cell proliferation, contact inhibition, invasion, Hippo signaling, and mitotic and neuro-developmental pathways. Our study underscores the pan-cancer potential of the non-coding transcriptome for identifying biomarkers and regulators of cancer progression.