OmiEmbed: A Unified Multi-task Deep Learning Framework for Multi-omics Data

Xiaoyu Zhang ([email protected])

Data Science Institute, Imperial College London

OmiEmbed is a unified framework for deep learning-based omics data analysis, which supports:

1. Multi-omics integration
2. Dimensionality reduction
3. Omics embedding learning
4. Tumour type classification
5. Phenotypic feature reconstruction
6. Survival prediction
7. Multi-task learning for aforementioned tasks

Paper Link: arXiv

• CPU or NVIDIA GPU + CUDA CuDNN
• Python 3.6+
• Python Package Manager
  • Anaconda 3 (recommended)
  • or pip 21.0+
• Python Packages
  • PyTorch 1.2+
  • TensorBoard 1.10+
  • Tables 3.6+
  • scikit-survival 0.6+
  • prefetch-generator 1.0+
• Git 2.7+

• Clone the repo

git clone https://github.com/zhangxiaoyu11/OmiEmbed.git
cd OmiEmbed

• Install the dependencies
  • For conda users

  conda env create -f environment.yml
  conda activate omiembed

  • For pip users

  pip install -r requirements.txt

• Train and test using the built-in sample dataset with the default settings

python train_test.py

• Check the output files

cd checkpoints/test/

• Visualise the metrics and losses

tensorboard --logdir=tb_log --bind_all

If you use this code in your research, please cite our paper.

@misc{zhang2021omiembed,
      title={OmiEmbed: a unified multi-task deep learning framework for multi-omics data}, 
      author={Xiaoyu Zhang and Yuting Xing and Kai Sun and Yike Guo},
      year={2021},
      eprint={2102.02669},
      archivePrefix={arXiv},
      primaryClass={q-bio.GN}
}

This source code is licensed under the MIT license.