sea-comet/CCL-Discovery

Source code for our paper: Discovery of Primary Prostate Cancer Biomarkers using Cross-Cancer Learning

Discovery of Primary Prostate Cancer Biomarkers using Cross-Cancer Learning

Introduction

This repository is for our submitted paper for Scientific Reports '[Discovery of Primary Prostate Cancer Biomarkers using Cross-Cancer Learning]'. The code is modified from DeePathology.

Installation

This repository is based on Tensorflow 2.2.0 For installing tensorflow, please follow the official instructions in here. The code is tested under Python 3.6 on Ubuntu 18.04.

Associate packages include: h5py, SHAP, sklearn.

Data

Our prepared data can be downloaded from CCL-Discovery(data). Put all files in this folder to data_process folder in the root directory.

Usage

1. Setup the parameters accordingly in option.py

2. Train the model for our autoencoder to obtain SHAP scores. Run:

   cd code
   python mlc-ae.py --phase train

3. Test the model of autoencoder and draw the SHAP visualization. Run:

   cd code
   python mlc-ae.py --phase test

4. Train the model for our evaluation classifier, in where we have attached sample score files. Run:

   cd code
   python eval-classifier.py --phase train

5. Test the model for our evaluation classifier. Run:

   cd code
   python eval-classifier.py --phase test