AltumAge

🧬 AltumAge: A Pan-Tissue DNA Methylation Epigenetic Clock Based on Deep Learning

AltumAge is a state-of-the-art epigenetic clock that predicts biological age from DNA methylation data across multiple tissue types. Built using deep learning, AltumAge demonstrates superior performance compared to traditional epigenetic clocks.

🎯 Key Features

• Pan-tissue compatibility: Works across multiple tissue types
• Deep learning architecture: Leverages neural networks for improved accuracy
• Multi-platform support: Compatible with Illumina 27k, 450k, and EPIC arrays
• PyTorch compatibility: Available in both TensorFlow and PyTorch formats
• Easy integration: Now available through the pyaging package

📊 Performance Highlights

• Trained, validated, and tested on 142 datasets
• Outperforms Horvath's 2013 model across multiple metrics
• Robust performance across diverse tissue types and age ranges

🚀 Quick Start

Option 1: Using pyaging (Recommended)

The easiest way to use AltumAge is through pyaging:

pip install pyaging

Then follow the DNA methylation age prediction tutorial.

Option 2: Standalone Usage

Prerequisites

pip install tensorflow==2.5.0 numpy pandas scikit-learn

Basic Usage

import tensorflow as tf
import numpy as np
import pandas as pd
from sklearn import linear_model, preprocessing

# Load CpG sites
cpgs = np.array(pd.read_pickle('example_dependencies/multi_platform_cpgs.pkl'))

# Load your methylation data
data = pd.read_pickle('example_dependencies/example_data.pkl')
methylation_data = data[cpgs]

# Load scaler and model
scaler = pd.read_pickle('example_dependencies/scaler.pkl')
AltumAge = tf.keras.models.load_model('example_dependencies/AltumAge.h5')

# Scale and predict
methylation_data_scaled = scaler.transform(methylation_data)
predicted_ages = AltumAge.predict(methylation_data_scaled).flatten()

📋 Detailed Instructions

1. Data Preparation

AltumAge requires:

• DNA methylation beta values from Illumina arrays (27k, 450k, or EPIC)
• Selection of 20,318 specific CpG sites (provided in CpGsites.csv)

2. Model Loading

# For TensorFlow users
AltumAge = tf.keras.models.load_model('example_dependencies/AltumAge.h5')

# For PyTorch users
import torch
AltumAge_pytorch = torch.load('dependencies/AltumAge.pt')

3. Preprocessing Pipeline

1. Select the required CpG sites in the correct order
2. Scale using the provided RobustScaler
3. Fill up missing values with 0 after scaling
4. Input to the model for age prediction

📁 Repository Structure

AltumAge/
├── example.ipynb                    # Complete usage example
├── example_dependencies/            # Required files for running AltumAge
│   ├── AltumAge.h5                 # TensorFlow model
│   ├── multi_platform_cpgs.pkl     # List of CpG sites
│   ├── scaler.pkl                  # Preprocessing scaler
│   └── example_data.pkl            # Example dataset
├── dependencies/
│   └── AltumAge.pt                 # PyTorch model
├── CpGsites.csv                    # Required CpG sites
└── supplementary_results/          # Detailed performance metrics

💾 Data Availability

Access our comprehensive dataset collection:

• Raw data from ArrayExpress and GEO
• Organized methylation data (non-normalized)
• Google Drive Repository

📚 Citation

If you use AltumAge in your research, please cite:

@article{de_Lima_Camillo_AltumAge,
    author = {de Lima Camillo, Lucas Paulo and Lapierre, Louis R and Singh, Ritambhara},
    title = {A pan-tissue DNA-methylation epigenetic clock based on deep learning},
    journal = {npj Aging},
    volume = {8},
    pages = {4},
    year = {2022},
    doi = {10.1038/s41514-022-00085-y},
    publisher = {Springer Nature},
    URL = {https://doi.org/10.1038/s41514-022-00085-y}
}

📧 Contact

For questions or collaborations, please contact:

• Lucas Paulo de Lima Camillo: lucas_camillo@alumni.brown.edu

📄 License

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