Abstract
Abstract
Purpose
Decisions in IDH-mutant lower-grade glioma rely on grade and MGMT promoter methylation, yet patients concordant for both show survival from under two years to over a decade. Existing transcriptomic methods assign discrete subtypes and give no per-sample score. We developed and validated a continuous one.
Methods
The Transcriptomic Ambiguity Index (TAMI) is the Shannon entropy of the posterior from a classifier separating glioblastoma from lower-grade glioma. A 200-gene panel was selected in The Cancer Genome Atlas (TCGA); values were rank-normalized within batch, scored, converted to entropy and standardized. IDH status came from PanCanAtlas molecular subtype, not histology. Validation used two Chinese Glioma Genome Atlas (CGGA) cohorts on separate platforms. Overall survival was analyzed by Cox regression with Benjamini-Hochberg false discovery rate (FDR) control at 0.05.
Results
In WHO grade 3 IDH-mutant glioma the hazard ratio per standard deviation (SD) of TAMI was 1.61 (1.25–2.06) in TCGA, 1.48 (1.23–1.79) in CGGA-693 and 1.93 (1.30–2.87) in CGGA-325, all surviving FDR correction. The association was stronger in molecularly confirmed than in histologically defined cases, so IDH-wildtype contamination does not explain it. In grade 2 disease it was null in two of three cohorts. Estimates were essentially unchanged after adjustment for 1p/19q codeletion, and TAMI remained independent in models containing codeletion, MGMT status, grade, age and treatment.
Conclusion
Classifier entropy carries prognostic information in grade 3 IDH-mutant glioma beyond grade, codeletion, MGMT status and treatment, and replicates across two platforms. It is not specific to grade 3 and requires standardization against the target cohort.