TULIP

The template-based modeling for accurate ligand-protein complex structure prediction (TULIP).

Input : Query protein, Ligand molecule (.smiles ), Protein Template Hits

Output : .sdf file of ligand molecule that potentially binds to input query protein

Setup Environment

1. The required packages are found in environment.yml

conda env create -f environment.yml

2. Once the installation is completed, activate the conda environment

conda activate tulip

3. The pipeline makes use of UCSF Chimera and PyRosetta. Please install them through their websites

UCSF Chimera Download , this link will take users to: https://www.cgl.ucsf.edu/chimera/download.html

PyRosetta Download , this link will take users to: https://www.pyrosetta.org/downloads

To adjust the target ligand's binding pose and orientation by rotation and translation, TULIP uses LS-align to align the target ligand with template ligands of higher similarity by both flexible and rigid alignments.

The LS-align tool can be downloaded and compiled into local machine from the below link:

LS-align , this link will take users to: https://zhanggroup.org/LS-align/

Run

To run TULIP pipeline run the following:

git clone https://github.com/BioinfoMachineLearning/tulip.git
cd ProteinLigand
python3 complex_generator.py --config=../configs/configs.yml

configs/configs.yml contains the configuration of the pipeline. Enter the correct paths in config file.

Post Processing

Once the ligand templates are identified using the above script, use the script below to run the post-processing for multi-atom ligands. Update the ls_align_path variable in the script to the path where LS-align is installed.

python3 post_process.py

For ions and small molecules run:

python3 process_small_mol.py

Clustering

To perform clustering in case we have same ligand SMILES for a protein.

python3 cluster_files.py

Outputs

TULIP returns the identified ligand molecules in ranks. Example : rank_0.sdf, rank_1.sdf, and so on for each protein.