Survival Queries
Examples of survival analysis using gene expression.
Basic Survival Analysis
Examine whether gene expression is associated with patient outcomes.
Prognostic Queries
Does high DDR1 expression predict worse survival in kidney cancer?What you’ll get:
- Kaplan-Meier survival curves (High vs Low expression)
- Log-rank test p-value
- Hazard ratio with confidence interval
- Median survival for each group
Association Queries
Is BRCA1 expression associated with overall survival in ovarian cancer?Tests whether expression level is related to survival outcomes.
Phrasing Variations
All of these work:
Does high DDR1 predict worse survival in KIRC?
Is DDR1 expression prognostic in kidney cancer?
Do patients with high DDR1 have worse outcomes in RCC?
Is DDR1 associated with survival in renal cell carcinoma?BioQuery understands various phrasings. If one doesn’t work, try rephrasing.
Cancer-Specific Examples
Kidney Cancer
Does high DDR1 expression predict worse survival in clear cell RCC?
Is VHL expression associated with survival in kidney cancer?
Do patients with high HIF1A have worse prognosis in KIRC?Breast Cancer
Does high Ki67 predict worse survival in breast cancer?
Is ESR1 expression prognostic in ER+ breast cancer?
Do patients with high HER2 expression have worse outcomes?Lung Cancer
Does high EGFR predict survival in lung adenocarcinoma?
Is KRAS expression associated with prognosis in NSCLC?
Do patients with high PD-L1 have better survival in lung cancer?Glioblastoma
Is MGMT expression associated with survival in glioblastoma?
Does high EGFR predict worse survival in GBM?
Do patients with high IDH1 have better prognosis in glioma?Ovarian Cancer
Does high BRCA1 predict survival in ovarian cancer?
Is CA125 expression associated with prognosis in ovarian tumors?
Do patients with high VEGFA have worse survival in OV?Understanding Results
Kaplan-Meier Curves
- Y-axis: Survival probability (1.0 = 100% alive)
- X-axis: Time (typically months or years)
- Two curves: High expression (red) vs Low expression (blue)
- Tick marks: Censored patients (still alive at last follow-up)
- Shading: 95% confidence interval
Hazard Ratio Interpretation
| HR | Interpretation |
|---|---|
| HR = 1 | No association |
| HR > 1 | High expression = higher risk of death |
| HR < 1 | High expression = lower risk of death |
| HR = 2 | 2x higher risk in high expression group |
Hazard ratios should be interpreted with confidence intervals. HR = 1.5 [0.8-2.8] is not statistically significant.
P-value
- p < 0.05: Statistically significant association
- p ≥ 0.05: No significant association detected
Survival Endpoints
Overall Survival (OS)
Does high MYC predict worse overall survival in lymphoma?Time from diagnosis to death from any cause.
Progression-Free Survival
Is EGFR associated with progression-free survival in lung cancer?Time from treatment to disease progression or death.
Not all TCGA cancers have progression-free survival data. If unavailable, overall survival is used.
Stratification
Median Split (Default)
By default, patients are split at median expression:
- High: Above median
- Low: Below or equal to median
Custom Cutoffs
Custom expression cutoffs (e.g., top quartile vs bottom quartile) are planned for a future release.
Limitations
Sample Size
Survival analysis requires sufficient events (deaths) to be meaningful:
- Small cancers (n < 100) may have wide confidence intervals
- Rare cancers may not have enough follow-up
Confounding Factors
TCGA data doesn’t control for:
- Treatment received
- Stage at diagnosis (partially available)
- Comorbidities
- Patient demographics
Observational Data
Survival associations are correlational, not causal. High expression correlating with poor survival doesn’t mean the gene causes poor outcomes.
Combining with Other Analyses
For comprehensive biomarker validation:
- Expression: “Is [gene] overexpressed in [cancer]?”
- Survival: “Does high [gene] predict worse survival?”
- Mutations: “Is [gene] frequently mutated?”
Genes that are overexpressed AND predict poor survival are stronger therapeutic target candidates.