IntegrAO: Integrate Any Omics

This is the official codebase for Integrate Any Omics: Towards genome-wide data integration for patient stratification.

     

Updates:

[2025.03.02] We added the functionalities of extracting feature importance for the unsupervised and supervised IntegrAO models! Feel free to check it out here: Unsupervised integration feature importance and Supervised integration feature importance. Welcome for suggestions!

[2025.01.23] 🥳 IntegrAO has been accepted to Nature Machine Intelligence!

[2024.01.15] 🥳 IntegrAO Preprint available!

🔨 Hardware requirements

IntegrAO package requires only a standard computer with enough RAM to support the in-memory operations.

🔨 Installation

IntegrAO works with Python >= 3.7. Please make sure you have the correct version of Python pre-installation.

1. Create a virtual environment: conda create -n integrAO python=3.10 -y and conda activate integrAO
2. Install Pytorch 2.1.0
3. IntegrAO is available on PyPI. To install IntegrAO, run the following command: pip install integrao

For developing, clone this repo with following commands:

$ git clone this-repo-url
$ cd IntegrAO
$ pip install -r requirement.txt

🧬 Introduction

High-throughput omics profiling advancements have greatly enhanced cancer patient stratification. However, incomplete data in multi-omics integration presents a significant challenge, as traditional methods like sample exclusion or imputation often compromise biological diversity and dependencies. Furthermore, the critical task of accurately classifying new patients with partial omics data into existing subtypes is commonly overlooked. We introduce IntegrAO, an unsupervised framework integrating incomplete multi-omics and classifying new biological samples. IntegrAO first combines partially overlapping patient graphs from diverse omics sources and utilizes graph neural networks to produce unified patient embeddings.

An overview of IntegrAO can be seen below.

📖 Tutorial

We offer the following tutorials for demonstration:

• NEW: Unsupervised integration feature importance
• NEW: Supervised integration feature importance
• Integrate simulated butterfly datasets
• Integrate simulated cancer omics datasets
• Classify new samples with incomplete omics datasets

Citing IntegrAO

@article{ma2025moving,
  title={Moving towards genome-wide data integration for patient stratification with Integrate Any Omics},
  author={Ma, Shihao and Zeng, Andy GX and Haibe-Kains, Benjamin and Goldenberg, Anna and Dick, John E and Wang, Bo},
  journal={Nature Machine Intelligence},
  volume={7},
  number={1},
  pages={29--42},
  year={2025},
  publisher={Nature Publishing Group}
}
}