EPIRNX is a deep neural network based on ResNeXt used to predict enhancer-promoter interactions. Its full name is Enhancer-Promoter Interactions predictor based on ResNeXt and written in Python and implemented using TensorFlow 2.
The DNA sequence data of EPIRNX is as same as EPIANN, the details of the data and the data augmentor (Data_Augmentation.R) can be seen in https://github.com/wgmao/EPIANN.
As for one-hot encoding, we supply a Python scripts called sequence_processing.py, you can see it at root directory, but first, you need to use Data_Augmentation.R to get the imbalanced/balanced sequence data file which is ended with .fasta.
You can use train.py to train your own models, the parameters are all have default value and explained in the following table.
| Script arguments | Default | Explanation |
|---|---|---|
| model-name | ResNeXt | The name of model. |
| dataset-dir | data/EPIs/ | Dataset root path. |
| dataset-type | imbalance | balance or imbalance dataset. |
| dataset-coding | onehot | Data encoding mode. onehot or embedding. |
| cell-line | GM12878 | The cell-line which you want to Train. |
| batch-size | 32 | The number of samples used to train in each epoch. |
| epochs | 25 | The number of training iterations. |
| train-verbose | 1 | The style of progress bar, 0 or 1 or 2, 0=silent, 1=progress bar, 2=one line per epoch. |
You can use trainsfer_train.py to train your own trainsfer models, the parameters are all have default value and explained in the following table.
| Script arguments | Default | Explanation |
|---|---|---|
| model-name | ResNeXt | The name of model. |
| dataset-dir | data/EPIs/ | Dataset root path. |
| dataset-type | imbalance | balance or imbalance dataset. |
| dataset-coding | onehot | Data encoding mode. onehot or embedding. |
| cell-line | GM12878 | The cell-line of the pre-train model. |
| batch-size | 64 | The number of samples used to train in each epoch. |
| epochs | 20 | The number of training iterations. |
| train-verbose | 1 | The style of progress bar, 0 or 1 or 2, 0=silent, 1=progress bar, 2=one line per epoch. |
You can use test.py to test models, the parameters are all have default value and explained in the following table.
| Script arguments | Default | Explanation |
|---|---|---|
| model-name | ResNeXt | The name of model. |
| dataset-dir | data/EPIs/ | Dataset root path. |
| dataset-type | imbalance | balance or imbalance dataset. |
| dataset-coding | onehot | Data encoding mode. onehot or embedding. |
| train-cell-line | GM12878 | The train-cell-line of the model. |
| batch-size | 64 | The number of samples used to train in each epoch. |
| test-verbose | 1 | The style of progress bar, 0 or 1 or 2, 0=silent, 1=progress bar, 2=one line per epoch. |
- Python==3.7.4
- tensorflow-gpu==2.1.0
- sklearn==0.23.2
- json==2.0.9
- numpy==1.18.5
- matplotlib==3.3.0