# TRACERx Expression Data

VST-normalized gene expression data from the TRACERx lung cancer cohort.

## Contents

- `tracerx_vst_counts.parquet`: VST-normalized expression matrix (Parquet format)
  - Rows: 20136 genes
  - Columns: 1051 samples
  - Size: 106.0 MB

## Usage

```python
from huggingface_hub import hf_hub_download
import pandas as pd

# Download the expression data (Parquet - faster)
local_path = hf_hub_download(
    repo_id="jruffle/tracerx_expression_data",
    filename="tracerx_vst_counts.parquet",
    repo_type="dataset"
)

# Load the data
expression_df = pd.read_parquet(local_path)
```

## Data Format

- Index: Gene symbols (gene_id)
- Columns: Sample IDs
- Values: VST-normalized expression values

## Why Parquet?

Parquet format provides:
- ~3x smaller file size than CSV
- ~10-100x faster loading times
- Native pandas support

## License

This data is provided for research purposes as part of the CRUK Datathon 2025.