LigTMap Target Prediction Method for Small Molecules

Welcome to the LigTMap target and activity prediction for small molecules. This method currently support prediction for 17 target classes including 6000+ protein targets. This code includes the main prediction workflow and all data/models so it can be run offline in your working computer. However, for better visualization of the prediction result, our web server is recommended.

Visit our online server at https://cbbio.online/LigTMap/

The code is still in its early stage. You are welcome to feedback or contribute in making in silico target prediction a truly powerful method for novel drug discovery for everyone!

Software requirements:

Anaconda, RDKit, Openbabel, MOPAC2016, ODDT, PSOVina, MGLTools, and Python libraries.

Specifically, our method has been tested with these versions:

• python 2.7 (from anaconda)
• rdkit-2016.03.4
• numpy-1.11.3
• openbabel-3.0.0
• pychem-1.0
• pybel-0.12.2
• scikit-learn-0.19.2
• scipy-1.1.0
• pandas-0.23.4
• boost-1.59.0

INSTALLATION

This code was tested on MacOS X 11.2, CentOS 7.6 and 7.8. We will be glad to know if it works also on your platform!

1. MOPAC2016

MOPAC2016 can be downloaded from http://openmopac.net/MOPAC2016.html You need a license to use, please go to the homepage to obtain a license. The license key will be emailed to you.

In essence, the installation steps are:

Create the directory:

% sudo mkdir -p /opt/mopac
% sudo chmod 777 /opt/mopac

Copy over the MOPAC executable and library that are obtained after unpacking the downloaded package:

% cp <source-path>/MOPAC2016.exe /opt/mopac
% cp <source-path>/libiomp5.so /opt/mopac
% chmod +x /opt/mopac/MOPAC2016.exe

Add the following lines to your .bashrc start-up script:

alias mopac='/opt/mopac/MOPAC2016.exe'
export LD_LIBRARY_PATH=/opt/mopac:$LD_LIBRARY_PATH

Source the start-up script, e.g.

% source ~/.bashrc

Install the license key that you have received in your email:

% /opt/mopac/MOPAC2016.exe <license-key>

Test the installation using the given example:

% mopac Example_data_set.mop

If the run is completed with output at Example_data_set.out, then your installation is successful!

2. Anaconda

Download and install the latest version of Anaconda from https://www.anaconda.com/download/. Simply run the Anaconda3-xxx.sh file and provide an installation directory, e.g.

% ./Anaconda3-2020.11-Linux-x86_64.sh
...
/home/user/opt/anaconda3

It's good to organize your program files in one central place like /home/user/opt/

3. Setup a Python 2.7 environment in your anaconda

% conda create -n ligtmap -c rmg rdkit python=2.7 
% conda activate ligtmap

After activation, your default python interpreter should be the one from the ligtmap env. Check to confirm:

% which python
e.g. /home/user/opt/anaconda3/envs/ligtmap/bin/python

4. Openbabel

Follow http://openbabel.org/wiki/Category:Installation to install Openbabel that suits your platform.

% conda install -c conda-forge openbabel  

5. PyChem

Download and install PyChem from https://code.google.com/archive/p/pychem/downloads

% tar cvfz pychem-1.0.tar.gz
% cd pychem-1.0
% python setup.py install

6. PyBel

% python -m pip install pybel

7. Scikit-learn + Pandas

% conda install -c conda-forge scikit-learn=0.19.2
% conda install -c conda-forge pandas=0.23.4 

8. ODDT

Make sure you have all previous libraries installed with the correct version before running this:

% python -m pip install oddt

Troubleshooting

In case you meet errors in between and want to remove and reinstall from Step 3:

% conda env remove --name ligtmap

9. PSOVina

Download and install boost-1.59.0.tar.gz from https://sourceforge.net/projects/boost/files/boost/1.59.0/ if boost is not yet in your system.

% tar xfz boost_1_59_0.tar.gz
% cd boost_1_59_0
% ./bootstrap.sh --prefix=/home/user/opt/boost-1.59.0
% ./b2 -j 4
% ./b2 install

Add boost to the library path in .bashrc

export LD_LIBRARY_PATH=$HOME/opt/boost-1.59.0/lib:$LD_LIBRARY_PATH

Once your boost is in place, download and install psovina-2.0.tar.gz from https://sourceforge.net/projects/psovina/

% tar xfz psovina-2.0.tar.gz
% cd psovina-2.0/build/<your-platform>/release

Modify Makefile to suit your system setting, specifically give the location of the boost, e.g.: BASE=/home/user/opt/boost-1.59.0

% make
% mkdir /home/user/opt/psovina-2.0
% cp psovina psovina_split /home/user/opt/psovina-2.0

Make it accessible by adding the location of the compiled psovina to the PATH in .bashrc

export PATH=/home/user/opt/psovina-2.0:$PATH

10. MGLTools

Download and install MGLTools of your platform from http://mgltools.scripps.edu/downloads

% tar xfz mgltools_x86_64Linux2_1.5.6.tar.gz
% mv mgltools_x86_64Linux2_1.5.6 /home/user/opt
% cd /home/user/opt/mgltools_x86_64Linux2_1.5.6
% ./install.sh

Following the instructions at the end of the installation to include some variables in your .bashrc file.

11. gsplit (for MacOS X only)

Install some GNU utilities via Homebrew, especially, we need gsplit as an alternative to the darwin split.

% brew install coreutils  

12. LigTMap

Download and unpack ligtmap-0.1.tar.gz. You can move the program directory to anywhere.

% tar xfz ligtmap-0.1
% mv ligtmap-0.1 /home/user/opt

13. Setting environment variables

Define necessary environment variables in the .bashrc start-up script file:

export LIGTMAP=/home/user/opt/ligtmap-0.1
export MGLTools=/home/user/opt/mgltools_x86_64Linux2_1.5.6/

Finally, source the script file.

% source ~/.bashrc

HOW TO RUN TARGET PREDICTION

1. Prepare your molecule(s) to be predicted in input.smi, e.g. our benchmark molecules for HIV. Make sure you don't leave any empty lines in the file:

c1ccccc1Oc(ccc2)c(c23)n(c(=O)[nH]3)CC
c1c(C)cc(C)cc1Oc(ccc2)c(c23)n(c(=O)[nH]3)CC
N#Cc(c1)cc(Cl)cc1Oc(ccc2)c(c23)n(c(=O)[nH]3)CC
N#Cc(c1)cc(Cl)cc1Oc(ccc2)c(c23)n(C)c(=O)[nH]3

2. Prepare the list of targets in target.lst. For a complete list of supported targets, refer to $LIGTMAP/target.lst.

HIV
HCV

3. Activate the condo environment

% condo activate ligtmap 

4. Run the prediction

% $LIGTMAP/predict

The run will generate two directories Input and Output. Input stores each molecule SMILES in a separate file:

input_00001, input_00002, ...

Output stores prediction results for each molecule separately in directories.

In case you have a previous run, the Input and Output directories will be backuped to Input.xxx and Output.xxx.

5. Examine prediction results

In the summary section, the target class for which target proteins have been identified for the query molecule is marked Complete, Otherwise Fail.

For a molecule Input_xxxxx, the top-ranked targets sorted by the LigTMapScore can be found in Output/Input_xxxxx/IFP_result.csv.

This file contain 9 columns of data of the identified targets:

1. PDB
2. Class
3. TargetName
4. LigandName
5. LigandSimilarityScore
6. BindingSimilarityScore
7. LigTMapScore
8. PredictedAffinity
9. DockingScore

The binding mode (PDB) of the molecule at the target protein can be found in the corresponding directory Output/Input_xxxxx/TargetName/Complex

Citation

Our method paper is currently under review:

Shaikh, Faraz; Tai, Hio Kuan; Desai, Nirali; Siu, Shirley (2020): LigTMap: Ligand and Structure-Based Target Identification and Activity Prediction for Small Molecular Compounds. ChemRxiv. Preprint. https://doi.org/10.26434/chemrxiv.12923474.v2

Contact

Developer: Faraz Shaikh (shaikh.faraz78@gmail.com), Giotto Tai (hiokuantai@gmail.com)

Project PI: Shirley Siu (siuwengin@gmail.com | shirleysiu@um.edu.mo | https://twitter.com/ShirleyWISiu)

Computational Biology and Bioinformatics Lab (CBBIO)

University of Macau