PST-PRNA: Prediction of RNA-Binding Sites Using Protein Surface Topography and Deep Learning.

Description

PST-PRNA is a method to decipher RNA binding sites on protein surface based on protein surface topography. To achieve this, PST-PRNA builds the topographies and applies deep learning methods to learn from these. For convenient use, please visit the web service www.zpliulab.cn/PSTPRNA. For the standalone offline version, please install and use as follows.

Standard alone Software prerequisites

• Conda Conda is recommended for environment management.
• Python (3.6).
• reduce (3.23). To add protons to proteins.
• MSMS (2.6.1). To compute the surface of proteins.
• PSI-Blast (2.6.0+) To generate PSSM.
• hhblits (3.3.0) To generate HMM.
• CD-HIT(4.8.1) To clustering protein sequences .
• DSSP To standardize secondary structure assignment.

Some important Python packages

• BioPython (1.78). To parse PDB files.
• Pytorch (1.7.1). pytorch with GPU version. Use to model, train, and evaluate the actual neural networks.
• scikit-learn (0.24.1).

Specific usage

1 Download and install the standard alone software listed above.

Change the paths of these executable file at default_config/bin_path.py.

2 Topography preparing

(1) The script 'protein.py' contains the class RBP which interates all procedures that are needed to convert a RBP to topographies.

(2) For each protein, it takes tens of minutes to calculate topographies. So we recommend using parallel computing tools, such as slurm. The bash script 'prepare_all.slurm' helps for extracting topographies in parallel cooperating with the python script 'prepare_all.py'.

(3) Users can also use 'prepare_all.py' all alone for preprocessing data. The files containing RBP_ids are in data/pdbid_chain. And the path of PDB_id lists should be specific the two 'prepare_all' scripts.

3 Training

To train an ab initio model, simply uses the script 'train.py'. Specific the RBPs list in default_config/dir_options:

python train.py

4 Predicting

To predict new RNA-binding sites: a. set the dir_opts['PDB_list_to_predict'] (in default_config) referring to the list file containing the PDB names (one name one line) b. move the PDB files to folder dir_opts['raw_pdb_dir']: then execute：

a. python prepare_all.py
b. python predict.py

License

PST-PRNA is released under an MIT License.

Reference

If you use this code, please use the bibtex entry in citation.bib.