Molecular and clinical heterogeneity critically hinders better treatment outcome for glioblastomas (GBMs); integrative analysis of genomic and epigenomic data may provide useful information for improving personalized medicine. By applying training-validation approach, we identified a novel hypomethylation signature comprising of three CpGs at non-CpG island (CGI) open sea regions for GBMs. The hypomethylation signature consistently predicted poor prognosis of GBMs in a series of discovery and validation datasets. It was demonstrated as an independent prognostic indicator, and showed interrelationships with known molecular marks such as MGMT promoter methylation status, and glioma CpG island methylator phenotype (G-CIMP) or IDH1 mutations. Bioinformatic analysis found that the hypomethylation signature was closely associated with the transcriptional status of an EGFR/VEGFA/ANXA1-centered gene network. The integrative molecular analysis finally revealed that the gene network defined two distinct clinically relevant molecular subtypes reminiscent of different immature neuroglial lineages in GBMs. The novel hypomethylation signature and relevant gene network may provide new insights into prognostic classification, molecular characterization, and treatment development for GBMs.
Posted in featured publications.