https://github.com/leekanggeun/CPEM

CPEM: Accurate cancer type classification based on somatic alterations using an ensemble of random forests and a deep neural network
https://github.com/leekanggeun/CPEM

Last synced: 3 months ago
JSON representation

CPEM: Accurate cancer type classification based on somatic alterations using an ensemble of random forests and a deep neural network

• Host: GitHub
• URL: https://github.com/leekanggeun/CPEM
• Owner: leekanggeun
• Created: 2019-01-14T03:52:51.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2021-07-07T01:46:31.000Z (almost 3 years ago)
• Last Synced: 2024-01-22T02:21:24.311Z (5 months ago)
• Language: Python
• Size: 123 KB
• Stars: 2
• Watchers: 2
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Lists

• awesome-biomedical-machine-learning - *** CPEM: Accurate cancer type classification based on somatic alterations using an ensemble of random forests and a deep neural network

README

        
## CPEM

**CPEM: Accurate cancer type classification based on somatic alterations using an ensemble of a random foerest and a deep neural network**

----------------------------------

The paper is available [here](https://www.nature.com/articles/s41598-019-53034-3)

![Overview](overview.PNG)

## DATA

Data is avaliable on below address because of data size.

You can download the data and lable in here 

[Somatic alteration information](https://drive.google.com/file/d/1uDFNhzsodQky71bqykmJGcWnBfhO5_hp/view?usp=sharing)

, [Cancer type information](https://drive.google.com/file/d/1l2uggi6rfbNwarkR2fcAsOf2q5CLfbyu/view?usp=sharing)

## Quick start

----------------------------------

We recommend the below argument to use the similar verification method as this paper.

```sh

python main.py --epoch=120 --batch_size=16 --lr=1e-3 --inner=10 --outer=10 --ensemble --search --feature_selection

```

#inner: This argument can be 'LOOCV' (Leave One Out Cross Validation) or 'K' (K should be integer) for K-fold inner cross validation.

#outer: This argument can be 'LOOCV' (Leave One Out Cross Validation) or 'K' (K should be integer) for K-fold outer cross validation.

#ensemble: Whether you use ensemble model

#search: Whether you search the optimized number of features

#feature_selection: Whether you execute feature selection

## Requirements

1. tensorflow-gpu>=2.1.0 

2. scikit-learn>=0.18.1 

3. tensorflow-addons == 0.9.0 

4. hdf5sotrage

## Future work

To increase the performance of feature selection, we plan to modify the feature selection code to Tensorflow.