PANDA: Prioritization of autism‐genes using network‐based deep‐learning approach

This repository is a re-implementation of PANDA: Prioritization of autism‐genes using network‐based deep‐learning approach in PyTorch.

Requirements

Code environment is managed via Anaconda in this project.

To create an environment and install all dependencies:

conda env create -f environment.yml

To clone and run this project locally:

git clone https://github.com/ChengyuanSha/PANDA

File Structure

• data folder:
  • HMIN_edgelist.csv: Human molecular interaction network (HMIN) in an edge list format
  • labeled_genes.csv: Graph nodes labels corresponding to HMIN
• src folder contains implementation codes:
  • tests folder: tests for quality assurance.
  • experiments folder: other testing experiments
  • GCN.py: definition of graph convolution network model
  • main.ipynb: the training and evaluation jupyter-notebook.
  • read_data.py: data preprocessing
  • baselines.ipynb: other algorithms to compare with

Training

Our model takes a small amount of time to train since the dataset is small. To train my model, run main.ipynb in src. The training code is under the Model training section. To save model, uncomment code in Save model section.

Evaluation

To evaluate my model, run main.ipynb in src. The evaluation code is under the Model Evaluation section.

Pre-trained Models

Our model trained on HMIN with labeled autism genes dataset. You can download pretrained model in: src/pretrained_model.pth.

Results

Our model (PANDA) achieves the following performance comparing with Random Forest (RF), support vector machine (SVM), Linear Genetic Programming(LGP) on HMIN dataset:
The baseline code can be found in src/baselines.ipynb with sections: Random Forest, SVM, LGP

Model name	Accuracy	Precision	Recall	F1 score
PANDA	88%	1.00	0.88	0.93
RF	86%	0.98	0.84	0.91
SVM	85%	0.99	0.85	0.92
LGP	86%	0.98	0.84	0.91

Contributing

If you'd like to contribute, or have any suggestions, you can open an issue on this GitHub repository. All contributions welcome! All content in this repository is licensed under the MIT license.

Reference

Zhang, Y., Chen, Y., & Hu, T. (2020). PANDA: Prioritization of autism‐genes using network‐based deep‐learning approach. Genetic epidemiology, 44(4), 382-394.