Digital Health Innovation and Informatics

PD 14 - Digital Health Information & Informatics - Poster Discussion

1122 - Hypoxic Gene Expression Defines Novel Tumor Subtypes in Glioblastoma

Tuesday, October 23
5:15 PM - 5:21 PM
Location: Room 217 A/B

Hypoxic Gene Expression Defines Novel Tumor Subtypes in Glioblastoma
D. B. Vanderbilt1,2, J. A. Vargo3, and J. M. Ruppert1,2; 1West Virginia University School of Medicine, Morgantown, WV, 2WVU Cancer Institute, Morgantown, WV, 3UPMC Hillman Cancer Center, Pittsburgh, PA

Purpose/Objective(s): Despite multimodal therapeutic advances, outcomes in glioblastoma multiforme (GBM) remain poor. Previous efforts in identifying GBM subtypes based on mRNA expression have been limited in clinical utility, as minimal survival difference between clusters was observed. In GBM and other tumor types, the role of hypoxia is increasingly recognized, and gene expression driven by hypoxia inducible factors (HIFs) is known to promote tumor recurrence and resistance to therapy. To better identify clinically meaningful correlates and hypothesis-generating patterns within the GBM transcriptome, we sought to apply unsupervised learning methods to characterize relationships amongst GBM specimens, with emphasis on hypoxic gene expression.

Materials/Methods: We performed clustering of GBM specimens by expression of known HIF target genes, using data from The Cancer Genome Atlas. Using automated modelling, we used Cox methods to screen for candidate hypoxic genes with survival significance. We then characterized the GBM subtypes with regard to hypoxic gene expression, patient survival, and other molecular and clinical parameters, and utilized established machine learning methods to predict cluster membership.

Results: We describe novel GBM clusters with significant differences in patient survival (p < 0.01; Cox regression), as well as distinct patterns of hypoxic gene expression. Of literature-documented HIF target genes, we identified 13 candidates as potentially significant with respect to patient survival. Collective visualization of these candidate genes revealed distinguishing patterns amongst the GBM hypoxic subtypes, and further investigation demonstrated robust variation in expression of certain hypoxic genes, including CA9, LOX, and IGFBP2. Accordingly, subtype prediction by machine learning methods, using the HIF target genes with putative survival significance as input features, had good specificity and sensitivity with respect to multiple algorithms including random forests. The cluster associated with best prognosis, in which median survival (737 days) was approximately two-fold greater than other cases (394 days; p < 0.01), exhibited substantially decreased specimen necrosis, despite the observation that necrosis had no overall association with survival.

Conclusion: Collectively, these data support the existence of GBM subtypes determined by hypoxic gene expression, demonstrate the relevance of these clusters with respect to patient survival, and provide distinct, distinguishing patterns of gene expression and histologic parameters amongst these clusters. This work also supports the feasibility of prospectively determining survival-significant subtype clustering of new GBM cases based on expression of a small number of HIF target genes.

Author Disclosure: D.B. Vanderbilt: None. J.A. Vargo: Honoraria; Brain Lab. J.M. Ruppert: Patent/License Fees/Copyright; West Virginia University.

Daniel Vanderbilt, MD, PhD

West Virginia University: Resident physician: Employee


Send Email for Daniel Vanderbilt


1122 - Hypoxic Gene Expression Defines Novel Tumor Subtypes in Glioblastoma

Attendees who have favorited this

Please enter your access key

The asset you are trying to access is locked. Please enter your access key to unlock.

Send Email for Hypoxic Gene Expression Defines Novel Tumor Subtypes in Glioblastoma