Abstract General Information

Title

EPIGENOMICS STEMNESS PREDICTION MODEL STRATIFIES IDHWT GLIOMAS BASED ON OVERALL SURVIVAL

Introduction, Objectives, Methods, Results, and Conclusion.

Stem cell phenotypes may underline glioma IDHwt, a poor prognosis cancer
subtype, therapeutic resistance due to the stemness properties as self-renewal and
proliferative potential. Although glioma molecular alterations have been described,
there are no prognosis distinctions noticed regarding IDHwt. Among genetic and
epigenetic classify technologies, DNA methylation seems to very well stratify brain
tumors, including gliomas. To quantify stemness, machine learning algorithms have
been applied in tumor samples and correlated with clinical and molecular data in
previous studies. We propose to define a novel metric to measure stemness in
tumors using a non-tumor induced Neural Stem Cell (iNSC) DNA methylation
signature. We used public DNA methylation data from 9 non-tumor samples to build
our prediction model. We then applied our model to classify 689 and 132 tumor
samples from the TCGA and the GLASS cohorts, respectively. To define our
signature and build our prediction model, we filtered probes by comparing iNSC with
normal brain tissue and mapped them to promoter/distal regions. Our model
stratified gliomas samples both by IDH mutation status and by molecular subtypes in
both TCGA and GLASS cohorts. IDHwt presented higher stemness indices
compared to IDHmut tumors. Interestingly, IDHwt tumors were stratified by the
stemness indices with significant differences in survival in both cohorts. The higher
the stemness index the poorer the overall survival after adjusting for age and
molecular subtype (TCGA: likelihood ratio test p < 0.001; GLASS: likelihood ratio
test p = 0.086). High stemness in gliomas IDHwt was found in previous work from
our group, but no difference in survival analysis had been found until now. Our
stemness prediction results indicated an enrichment of stemness features in glioma,
which is associated with prognosis. Elucidating how stemness plays a role in
high-grade gliomas could improve patient prognosis and tumor aggressiveness
understanding.

Keywords (separated by comma on a single line)

Glioma, Stemness, IDH, DNA methylation, machine learning, bioinformatics