Identifying piRNA targets on mRNAs in C. elegans using a deep multi-head attention network

Related paper:

Tzu-Hsien Yang, Sheng-Cian Shiue, Kuan-Yu Chen, Yan-Yuan Tseng and Wei-Sheng Wu*, "Identifying piRNA targets on mRNAs in C. elegans using a deep multi-head attention network". (Submitting)

Available Datasets:

1. The positive and the negative sets from the wild-type CLASH data
2. The positive and the negative sets from CSR-1 depleted CLASH data

Model Codes for the Proposed Method

Suggested running environments: Linux Ubuntu 14.04, Python 3.8.0

1. Download the codes from the following link: Download.
2. Unzip the file:
   unzip predict_code.zip
3. Change the working directory into the model codes:
   cd predict_code
4. Install the necessary packages:
   pip install -r requirements.txt
5. Prepare the piRNA-mRNA sequence pairs in the folder name "Input".
   Multiple piRNA-mRNA pairs are allowed to be in the same input file. The input format should followed the following examples:
   <piRNA sequence>,<mRNA sequence>
   Note: The input piRNA sequences and mRNA sequences should all start from the 5' end.
   TGGTGGATCGTCATTTGGTGG,CGAACACCCACCACCTTGTGAACCGCCTTGT
   TTCTCATCCGGTCCAAGAGGT,TAGACCTCGACGGACTTACTGGCGACCTTCC
   
6. Predict the binding probability of the given piRNA sequences and mRNA sequence segments:
   python main.py --input_file_name <Input_file_name>
7. Output: Probability of the binding events for each pairs will be written to 'Output_<Input_file_name>.csv'.