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

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