# mRPI Issue in Deep Learning

The dataset link file contains the links for the following:

- Training-validation set
- Random test set
- RBP-aware test set
- Reserved test set
- eCLIP independent test set
- List of RBPs collected from the CLIP-seq experiments


## Related paper:

Yu-Huai Yu+, Han-Ting Hong+, Tzu-Hsien Yang*, "Generalizable deep-learning-based mRNA-protein interaction prediction strongly depends on protein diversity", (submitting).

+: These authors contributed equally.

## Prepare the Environment

Suggested running environments: Linux Ubuntu 16.04.6, Python 3.8.13

We recommend that you can use the conda package to create a new environment. This will automatically install the required python packages. 

Here is an example: 

1. Install the Conda package for you system. The installation of the package can be found <a href="https://docs.conda.io/projects/conda/en/latest/user-guide/install/index.html">here</a>. 

2. Create the RPIPred Conda environment. This may take a while, depending on the network status.

```
conda create -n "RPIPred" python=3.8.13
```

3. Activate your RPIPred Conda environment. 

```
conda activate RPIPred
```

## Steps to Use RPIPred

1. Download the codes from the following link and unzip the file. Please skip it if you have done this step.

```
wget https://cobis-fs.bme.ncku.edu.tw/RPIPred/RPIPred.tar.gz
```

2. Unzip the file.

```
tar -zxvf RPIPred.tar.gz
```

3. Change the working directory.

```
cd RPIPred
```

4. If this is the first time you use RPIPred, run the following command to install necessary packages. 

```
pip install -r requirements.txt
```

5. Prepare Input RNA–Protein Pairs input files.



The input RNA-protein pairs (.csv) should be in the format of `RNA_Seq,protein_Seq`. If the selected mode is "3D", users additionally **need to provide an csv file for PDB file paths**. Please refer to the example input file for more details.

6. Predict the probability.

```
python main.py -i <input_csv_path> --pdb_path_file <pdb_csv_path> -o <output_csv_path> -m <onehot|proteinbert|3D>
```

>**Required arguments:**
>
>* -i: The input csv file path containing RNA–RBP pairs.
>
>* --pdb_path_file: The input csv file containing PDB file paths for the RBPs (required only for 3D mode).
>
>* -o: The output csv file path containing the predicted probabilities of RNA–protein interactions. 
>
>The format of the output csv file will be as follows: `RNA_Seq,protein_Seq,predicted_probability`. Please refer to the example output file for more details.
>
>**Optional arguments:**
>
>* -m: The encoding mode of protein for RPIPred. The default mode is "onehot". You can also choose to use "proteinbert" or "3D" as the encoding mode.

## Example Output Results
If we use the following as our inputs with the example command:

example_input.csv
![](images/input_1.png)
PDB.csv
![](images/input_2.png)

```
python main.py -i data/example_input.csv -o output/output.csv -m onehot
python main.py -i data/example_input.csv -o output/output.csv -m proteinbert
python main.py -i data/example_input.csv --pdb_path_file data/PDB.csv -o output/output.csv -m 3D
```
output.csv
![](images/output.png)

Output format explanation: 

The output file contains three columns: `RNA_Seq,protein_Seq,predicted_probability`. The predicted_probability is the probability that the RNA–protein pair interacts.