---

name: tcga-survival-analysis description: Perform survival analysis on TCGA cancer data using cBioPortal API and lifelines. Use when user wants to analyze gene expression and survival outcomes, create Kaplan-Meier curves, run Cox proportional hazards models, or study specific genes (like TP53, KRAS, EGFR) in cancer cohorts (LUAD, LUSC, BRCA, etc.). Triggers on keywords like "TCGA", "survival analysis", "Kaplan-Meier", "hazard ratio", or cancer type abbreviations.

TCGA Survival Analysis

Perform survival analysis on TCGA data using cBioPortal API + lifelines.

Step 1: Environment Check

Before starting the analysis, verify that required dependencies are available:

# Check required packages
python -c "import pandas, numpy, matplotlib, requests, lifelines; print('All dependencies OK')"

Dependency check strategy (by priority):

1. Check the currently activated environment (or user-specified environment)
2. Missing packages → Install with uv pip install first
3. If uv unavailable → Use pip install
4. If installation fails or conflicts → Only then create a new conda environment

Required dependencies:

• pandas
• numpy
• matplotlib
• requests
• lifelines

To install missing packages:

# Prefer uv (faster)
uv pip install lifelines

# Or use pip
pip install lifelines

Step 2: Quick Start

import pandas as pd
import requests
from lifelines import KaplanMeierFitter
from lifelines.statistics import logrank_test

# cBioPortal API
CBIOPORTAL_API = "https://www.cbioportal.org/api"
study_id = "luad_tcga_pan_can_atlas_2018"  # LUAD Pan-Cancer Atlas

# Fetch EGFR expression data
# ... (see workflow.md for details)

# Perform Kaplan-Meier analysis
kmf = KaplanMeierFitter()
kmf.fit(time, event, label='Gene High')
kmf.plot_survival_function()

Workflow

See workflow.md for detailed workflow, including:

1. Environment check - Verify dependencies, install as needed
2. Data download - Fetch expression and clinical data from cBioPortal
3. Data processing - Merge expression and survival data
4. Survival analysis - Kaplan-Meier analysis and Log-rank test
5. Visualization - Survival curves and expression distribution plots

Key Functions

Module	Function
requests	cBioPortal API data retrieval
lifelines	Kaplan-Meier and Cox regression
pandas	Data processing and merging

Supported Cancer Types

Common cBioPortal TCGA study IDs:

• luad_tcga_pan_can_atlas_2018 - Lung Adenocarcinoma (LUAD)
• lusc_tcga_pan_can_atlas_2018 - Lung Squamous Cell Carcinoma (LUSC)
• brca_tcga_pan_can_atlas_2018 - Breast Cancer (BRCA)
• coadread_tcga_pan_can_atlas_2018 - Colorectal Cancer

Troubleshooting

• Network issues: cBioPortal API requires stable network connection
• Gene names: Use Entrez Gene ID (e.g., EGFR = 1956)
• Survival data: Check if OS_MONTHS and OS_STATUS columns exist