ProbAge: Probabilistic Inference of Epigenetic Age Acceleration

Overview

ProbAge is a Python package implementing the probabilistic model for epigenetic age acceleration analysis described in "Probabilistic inference of epigenetic age acceleration from cellular dynamics" (Nature Aging, 2024). This tool provides a mechanistic approach to understanding methylation dynamics and biological aging, addressing limitations in traditional epigenetic clock methods.

Model Architecture

Key Features

• Probabilistic modeling of methylation transitions at the cellular level
• Distinction between acceleration and bias components in methylation dynamics
• Robust inference framework tested on large-scale cohort data
• Compatible with both local execution and cloud-based analysis

Installation

Local Installation

1. Clone the repository:

git clone https://github.com/zuberek/probage.git
cd probage

2. Create the conda environment:

conda env create -f env/probage.yml
conda activate probage

Usage

Local Analysis

Use any_cohort_run_no_train.py or any_cohort_run_retrain.py to analyze your cohort data locally, either by reusing a pretrained model in blood (fast) or refitting a model from scratch (slower but suited when analysing tissues other than blood).

1. Modify the data path in any_cohort_run_no_train.py or any_cohort_run_retrain.py to point to your methylation data.
2. Set global model parameters parameters.
3. Run the analysis:

python any_cohort_run.py

Cloud-Based Analysis

Access our web interface at https://probage.streamlit.app/ for cloud-based analysis.

Data Format

Your input data should include two components:

1. Methylation Data Matrix

• Format: Rows as CpG sites, columns as samples
• Values: Beta values (0-1 range) representing methylation levels
• File format: CSV or similar tabular format

2. Metadata File

Required columns:

• age: Numerical values representing the chronological age of each sample
• control: Boolean values (True/False) indicating which samples should be used for batch correction
  • Set to True for samples that represent the "normal" or "healthy" population
  • Set to False for samples from the group being studied (e.g., disease group)
  • Example: When studying disease acceleration, set control=True for healthy participants and control=False for disease participants
  • Example: When studying a cohort of healthy participants, set control=True for all participants.

Additional columns are encouraged and can be included for further analysis. Common examples:

• sex: Biological sex of the participant
• smoking_status: Smoking history
• bmi: Body Mass Index
• disease_status: Specific conditions or phenotypes
• Any other sample-level information relevant to your analysis

Example metadata format:

sample_id,age,control,sex,smoking_status,treatment_group
SAMPLE_001,45,True,F,Never,Placebo
SAMPLE_002,52,True,M,Former,Placebo
SAMPLE_003,48,False,F,Current,Treatment

Note: The batch correction process uses only samples marked as control=True to establish the baseline methylation dynamics, ensuring that the acceleration estimates are not confounded by the condition being studied.

Citation

If you use ProbAge in your research, please cite:

@article{dabrowski2024probabilistic,
    title={Probabilistic inference of epigenetic age acceleration from cellular dynamics},
    author={Dabrowski, Jan K. and Yang, Emma J. and Crofts, Samuel J. C. and Hillary, Robert F. and Simpson, Daniel J. and McCartney, Daniel L. and Marioni, Riccardo E. and Kirschner, Kristina and Latorre-Crespo, Eric and Chandra, Tamir},
    journal={Nature Aging},
    volume={4},
    pages={1493--1507},
    year={2024},
    publisher={Nature Publishing Group}
}

Model Details

ProbAge introduces two key components for analyzing epigenetic aging:

1. Acceleration: Measures the proportional increase in methylation transition speeds across CpG sites
2. Bias: Quantifies global changes in methylation levels

These components provide mechanistic insights into aging-related methylation dynamics and help control for confounding factors in traditional epigenetic clock analyses.

Contributing

We welcome contributions! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

For questions and support:

• Create an issue on GitHub
• Contact the authors through the corresponding author information in the paper

Acknowledgments

This work was developped using data from the Generation Scotland study.