Datasets:

Cong123779
/

gbm_survival_data

language: en
license: mit
task_categories:
  - tabular-classification
  - feature-extraction
tags:
  - bioinformatics
  - cancer-research
  - glioblastoma
  - survival-prediction
  - gene-expression
pretty_name: GBM Survival Gene Expression Dataset
size_categories:
  - 1GB<n<10GB

🧬 GBM Survival Gene Expression Dataset

This dataset contains large-scale gene expression data (RNA-Seq and Microarray) curated for predicting the survival outcomes of patients with Glioblastoma Multiforme (GBM) and Lower-Grade Glioma (LGG).

It is specifically designed to be used with the GBM Survival Mamba project.

📂 Data Structure

The data is organized into three main phases:

1. `01_raw/` (Raw Data)

Untouched genomic data from international portals:

TCGA: RNA-Seq data from The Cancer Genome Atlas (USA).
CGGA: Chinese Glioma Genome Atlas data (Asia).
GEO: Various cohorts from NCBI Gene Expression Omnibus (REMBRANDT, GSE4412, etc.).

2. `02_processed/` (Cleaned & Normalized)

Ready-to-use data for machine learning:

training/: Internal Train/Val/Test splits used for model development.
validation/: External independent cohorts for performance verification.
Files include X.csv (feature matrices with 16,383 genes) and y.csv (survival time and status).

3. `03_metadata/`

Mapping tables between Gene Symbols and Probe IDs, as well as clinical metadata.

🚀 How to use with the Project

If you have cloned the GitHub repository, you can automatically sync this data by running:

python scripts/download_data_hf.py

This will download all files and place them in the correct data/ directory structure.

📝 Data Sources & Citations

If you use this data, please cite the original sources:

TCGA Research Network: https://www.cancer.gov/tcga
CGGA Database: http://www.cgga.org.cn/
NCBI GEO: https://www.ncbi.nlm.nih.gov/geo/

Project Maintainer: Cong, K. X.