An implementation of multimodal deep neural network, a new model for human breast cancer prognosis prediction.
Our manuscipt titled with "A multimodal deep neural network for human breast cancer prognosis prediction by integrating multi-dimensional data" has been accepted by IEEE/ACM Transactions on Computational Biology and Bioinformatics. If you find MDNNMD useful in your research, please consider citing:
Sun, D., Wang, M., & Li, A. (2018). A multimodal deep neural network for human breast cancer prognosis prediction by integrating multi-dimensional data. IEEE/ACM Transactions on Computational Biology and Bioinformatics.
[python 2.7](https://www.python.org/downloads/)
[TensorFilow 1.0](https://www.tensorflow.org/install/)
[scikit-learn 0.18](http://scikit-learn.org/stable/)
[cuda 8.0](https://developer.nvidia.com/cuda-downloads)
python MDNNMD.py
The Parameters of MDNNMD are in our configuration file mdnnmd.conf. The descriptions of these parameters of MDNNMD are provided below:
=================================================================================================
| PARAMETER NAME | DESCRIPTION |
=================================================================================================
| α,β,γ |α,β,γ are three damping factors used to balance the contribution for |
| |each DNN model. Here the sum of three damping factors should be equal 1.|
-------------------------------------------------------------------------------------------------
| K |the number of fold with cross validation experiment or an index file. |
-------------------------------------------------------------------------------------------------
| D1 |the data file of gene expression profile |
-------------------------------------------------------------------------------------------------
| D2 |the data file of copy number alteration profile. |
-------------------------------------------------------------------------------------------------
| D3 |the data file of clinical information |
-------------------------------------------------------------------------------------------------
| LABEL |the predict label of breast cancer patients with 1 or 0. |
-------------------------------------------------------------------------------------------------
| batch_size |mini-batch size. |
-------------------------------------------------------------------------------------------------
| bne |batch normalization epsilon. |
-------------------------------------------------------------------------------------------------
| active_function |active_function in our MDNNMD model, choose tanh or relu. |
-------------------------------------------------------------------------------------------------
The descriptions of output files of MDNNMD are provided below:
====================================================================================================================
| VARIABLE NAME | DESCRIPTION |
====================================================================================================================
| Prediction_score.txt |The final prediction score of all samples with 10 fold cross validation experiment. |
| |The output of the MDNNMD with a softmax function. |
--------------------------------------------------------------------------------------------------------------------
| Prediction_labels.txt |The prediction labels represent long-term patients with 0 and short-term patients with 1. |
--------------------------------------------------------------------------------------------------------------------
Author: Dongdong Sun @HILAB
Maintainer: Dongdong Sun
Mail: sddchina@mail.ustc.edu.cn
Date: 2017-5-30
Health Informatics Lab, School of Information Science and Technology, University of Science and Technology of China