Glioblastoma multiforme GBM
Of 60,498 genes, 1,816 are higher in the tumor than in normal brain and in whole blood, and are not made by blood immune cells. 669 are protein-coding, and 667 of those have supporting plasma Cell-free RNA (cfRNA)RNA fragments that circulate in blood plasma outside cells. Most come from blood cells; a small share comes from other tissues, including tumors. evidence; the rest are mostly non-coding genes the plasma sources cannot check.
Samples compared
- Tumor
- 153
- TCGA primary tumors
- Normal tissue
- 1,157
- 5 TCGA tumor-adjacent, 1,152 GTEx brain
- Whole blood
- 337
- GTEx samples from 328 donors
- Survival
- 152
- patients, 120 events (Survival endpointsCurated TCGA outcomes (Liu et al. 2018): OS is overall survival, PFI the progression-free interval and DSS disease-specific survival. LB2 uses whichever has the most events in each cancer type.)
Results in brief
- 1,816
- candidate genes: 667 with plasma cfRNA evidence, 9 protein-coding without, 1,140 non-coding
- 42
- genes in the tissue panel, Held-out AUCHow well an elastic-net model separates tumor from normal tissue, measured with nested cross-validation on samples the model did not train on. It describes tissue, not the accuracy of a blood test. 1.000
- 0
- genes associated with overall survival (FDR < 0.05, 150 tested)
Candidates
Filter by plasma evidence or discrimination, find a gene, and select any point or row for its full record. The table holds all 1,816 candidates.
How the candidates were selected
60,498 genes tested; 6,065 higher in tumor than in normal brain; 3,208 also higher than in whole blood; 1,816 remain after removing genes expressed in blood immune cells; 667 of these have plasma cfRNA evidence, 9 protein-coding candidates have none, and 1,140 are non-coding genes the plasma sources cannot list.
Each comparison uses a two-sided Wilcoxon rank-sum test. A gene passes when its q-value is below 0.05, its log2 fold change is at least 1, and the lower bound of the 95% confidence interval for its AUC is at least 0.70.
Genes made by any of 18 sorted blood immune cell types above 1 nTPM are then removed, because blood cells supply most plasma RNA. Genes missing from that reference, mostly non-coding, are kept.
Plasma evidence only reorders the list. A candidate not yet seen in plasma stays in, because every dataset misses genes. Read the full methods.
Tissue classifier panel
An elastic-net logistic regression was trained on the 300 top-ranked candidates to tell tumor from normal tissue. Its Held-out AUCHow well an elastic-net model separates tumor from normal tissue, measured with nested cross-validation on samples the model did not train on. It describes tissue, not the accuracy of a blood test. is 1.000, from nested five-fold cross-validation.
Before training, the TCGA-versus-GTEx offset was removed from the expression matrix with the tumor/normal contrast protected, as in the tissue comparison. That correction is fitted once on all samples, so the held-out AUC is not fully independent of it.
This number describes tumor and normal tissue, where separation is expected to be near perfect. It is not the accuracy of a blood test: that has to be measured in plasma from patients and controls.
Association with survival
None of the 150 top-ranked candidates is associated with overall survival at FDR < 0.05 (152 patients, 120 events). A null result here is informative: these genes mark the presence of the cancer, not necessarily its course.
For the 150 top-ranked candidates, a Cox model relates tumor expression, as a continuous value, to overall survival in 152 patients (120 events). No high/low cutpoint is searched for, since optimized cutpoints inflate false positives.
A Hazard ratio per SDFrom a Cox model with the gene’s tumor expression as a continuous variable: the change in hazard for each one-standard-deviation increase. Above 1, higher expression goes with a shorter time to the event; below 1, with a longer one. above 1 means higher expression goes with a shorter overall survival. These are associations in tissue, not evidence that a blood level predicts outcome.
Caveats for GBM
- Only 5 TCGA adjacent samples anchor the correction for study differences between TCGA and GTEx normals.
- The blood comparison sets TCGA tumors against GTEx blood. Study and biology cannot be separated there, which is why genes made by blood immune cells are removed as well.