GRNimmuneClock

Cell-Type Specific Aging Clocks for Immune Cells

GRNimmuneClock provides pre-trained aging clocks for immune cell types, trained on genes significantly associated with age. Predict biological age from gene expression data with cell-type specific models trained on multiple cohorts. Each clock also ships with a consensus gene regulatory network (GRN) for the same cell type, usable to interpret the clock's genes in terms of transcription factor (TF) activity.

Features

• 🔬 Cell-Type Specific: Separate models for CD4T and CD8T cells
• 🧬 Age-Associated Features: Trained on genes significantly correlated with age, not a fixed gene panel
• 🔗 Network Analysis: Access bundled consensus GRNs for TF-target exploration, and score TF activity from a clock's coefficients (tf_activity_from_coefs)
• 🔍 Interpretation: Out-of-sample permutation gene importance (permutation_gene_importance)
• 🎨 Visualization Tools: Built-in plotting functions for analysis
• 🚀 Easy to Use: Simple Python API
• 🔧 Training Pipeline: Tools to train custom aging clocks

Installation

pip install grnimmuneclock

Or install from source:

git clone https://github.com/janursa/GRNimmuneClock.git
cd GRNimmuneClock
pip install -e .

Quick Start

from grnimmuneclock import AgingClock, load_example_data
import grnimmuneclock.plotting as gplot

# Load pre-trained clock for CD4T cells
clock = AgingClock(cell_type='CD4T')

# Load example data
adata = load_example_data()

# Predict biological age
adata_predicted = clock.predict(adata)
print(adata_predicted.obs['predicted_age'])

# Visualize predictions
gplot.plot_predicted_vs_actual(adata_predicted, hue='sex')

See the tutorial.ipynb for more.

Supported Cell Types

• CD4T: CD4+ T cells
• CD8T: CD8+ T cells

Model Information

All models are:

• Algorithm: Ridge regression with StandardScaler
• Features: Gene expression values, restricted per cell type to genes significantly associated with age (not a GRN target list)
• Training: Multiple cohorts (European, Korean, Japanese, Chinese)
• Age Range: 20-80 years
• Species: Human
• Tissue: Peripheral blood

Per-model feature counts and held-out performance are in grnimmuneclock/models/<cell_type>/metadata.json.

Citation

If you use GRNimmuneClock in your research, please cite:

@article{nourisa2025grnimmuneclock,
  title={TBD},
  author={Nourisa, Jalil and others},
  journal={TBD},
  year={2025}
}

License

MIT License - see LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Support

For questions and issues, please open an issue on GitHub.