Abstract (Expand)
Glioma is a clinically and biologically diverse disease. It challenges diagnosis and prognosis due to its molecular heterogeneity and diverse regimes of biological dysfunctions which are driven by … genetic and epigenetic mechanisms. We discover the functional impact of sets of DNA methylation marker genes in the context of brain cancer subtypes as an exemplary approach how bioinformatics and particularly machine learning using self organizing maps (SOM) complements modern high-throughput genomic technologies. DNA methylation changes in gliomas comprise both, hyper- and hypomethylation in a subtype specific fashion. We compared pediatric (2 subtypes) and adult (4) glioblastoma and non-neoplastic brain. The functional impact of differential methylation marker sets is discovered in terms of gene set analysis which comprises a large collection of markers related to biological processes, literature data on gliomas and also chromatin states of the healthy brain. DNA methylation signature genes from alternative studies well agree with our signatures. SOM mapping of gene sets robustly identifies similarities between different marker sets even under conditions of noisy compositions. Mapping of previous sets of glioma markers reveals high redundancy and mixtures of subtypes in the reference cohorts. Consideration of the regulatory level of DNA methylation is inevitable for understanding cancer genesis and progression. It provides suited markers for diagnosis of glioma subtypes and disentangles tumor heterogeneity.
Authors: E. Willscher, H. Loffler-Wirth, H. Binder, Lydia Hopp
Date Published: 2015
Publication Type: Not specified
Human Diseases: brain glioma
https://orcid.org/0000-0002-3662-1023
