FRASE-bot-RDKit

FRASE-bot-RDKit is a collection of cheminformatics protocols designed for data processing on FRASE screening, developed in Python with RDKit, a powerful cheminformatics toolkit. This version eliminates the need for a Pipeline Pilot commercial license and provides flexibility for users to use the capabilities of RDKit in their cheminformatics workflows.

Features

• Python and RDKit: FRASE-bot-RDKit made use of the capabilities of RDKit to perform a variety of cheminformatics tasks in a Python environment.

• Elimination of Pipeline Pilot Requirement: Unlike the FRASE-bot-Pipeline-Pilot, FRASE-bot-RDKit operates independently of Pipeline Pilot, allowing users to run protocols without the need for a commercial license.

• Two Protocols: The workflow is divided into two distinct protocols, one for FRASE-bot screening processing, and the other for subsequent neural network determined for final screening.

• Efficiency: Current scripts including bash and python scripts are specifically designed for optimal performance in a Linux environment. This ensures efficient execution of calculations.

System Requirements

To use FRASE-bot-RDKit, ensure that you have the following prerequisites:

• Python: Install Python on your Linux system. The code is compatible with Python 3.x.

• Conda: Install miniconda3 on your Linux system.

• RDKit: Install the RDKit cheminformatics toolkit by miniconda3. Detailed information on RDKit can be found on the RDKit website.

• Tensorflow: Install Tensorflow by minconda3. More information can be found on the Keras website

Installation

We strongly recommend installing RDKit and tensorflow on Linux using conda. Alternatively, you can run the scripts on Jupyter Notebook.

1. To install RDKit cheminformatics toolkit, it's recommended to create a new conda environment for RDKit.

conda create --name rdkit-tools python=3.8

rdkit-tools can be changed to your favorable name.

2. Activate the rdkit enviroment:

conda activate rdkit-tools

3. Install RDKit:

conda install -c conda-forge rdkit

4. To install Tensorflow, create another new conda enviroment name

conda create --name Tensorflow python=3.8

Tensorflow can be changed to your favorable name.

5. Activate the tensorflow enviroment:

conda activate Tensorflow

6. Install tensorflow:

conda install -c conda-forge tensorflow

Workflow

1. FRASE database containing a total of 10,598 FRASEs is stored in the compressed file FRASE_database_PDB0.zip, utilized for FRASE-bot screening. First of all, save this file on your working directory and then proceed to decompress it using the command:

   unzip FRASE_database_PDB0.zip
   

2. Execute the main script:

   submit FRASE-screening.sh.
   

   This bash script, FRASE-bot-RDKit-batch.sh, will use python script FRASE-bot-RDKit.py to perform batch FRASE screening based on 10,598 FARASEs. Note that certain High-Performance Computing (HPC) systems may have limitations on job submission numbers. Adjust the submission array accordingly in the bash script FRASE-bot-RDKit-batch.sh: ##SBATCH --array 0-10598. Replace the third argument, CIB1-Ana-prot.pdb, with the name of your pdb file in the following command python FRASE-bot-RDKit.py $inp $folder_id CIB1-Ana-prot.pdb. In addition, modify the path of FRASE_database_PDB{folder_id}/ directory in the script FRASE-bot-RDKit.py: frase_db = f'/your_path/FRASE_database_PDB{folder_id}/{frase_inp}'

3. After execution, generate new files for subsequent machine learning using the following bash script.

   submit generate_IF_target.sh
   

4. Employ the FRASE-bot-RDKit-NN.sh script in the final step to generate screening outcomes.

   submit FRASE-bot-RDKit-NN.sh
   

   The interaction fingerprint outcomes of 10,598 FRASEs is stored in the file IF_FRASEdb.json, which was utilized as training data. Save this file to your working directory and specify its actual path in the script FRASE-bot-RDKit-NN.py: FRASEdb_file = '/your_path/IF_FRASEdb.json'

5. Final screened ligand fragments can be found under new generated folder screen_ligFragments/.

Examples

• Three target examples, represented by the structures 1bl7, 3wd9, and CIB1-Ana-prot, are available in the "Examples" directory. Navigate to the respective target's directory to access the screened ligand fragments, which are stored in the screen_ligFragments/ subdirectory for each example.