Active Site Modeling and Clustering (ASMC)

ASMC combines (i) homology modeling of family members (MODELLER), (ii) ligand-binding pocket search (P2RANK), (iii) structural alignment of modeled active sites (USalign) and (iv) density-based spatial clustering of obtained alignments (DBSCAN) in a single command line. Clustering step can be carried out on either structural or sequence alignment.

Installation

Installation with conda and pip

Download the latest GitHub release to obtain the code (https://github.com/labgem/ASMC/releases) and extract the code from the archive.

Then, use the following commands from the ASMC/ directory:

conda env create -n asmc -f env.yml
pip install ./

Conda will install all the python dependencies and two required third-party softwares:

• MODELLER (you still need to request the license key)
• USalign

The pip command is required to create the asmc command and use ASMC.

It's also possible to use only the pip install ./ command, but this will not install any third party software.

Third party Software dependencies

• P2RANK - for ligand-binding pocket detection (https://github.com/rdk/p2rank)
• MODELLER - for homology modeling (https://salilab.org/modeller/)
• USalign - for structural alignment (https://github.com/pylelab/USalign)

P2RANK setup

Download the p2rank tar.gz file (e.g: p2rank_2.5.tar.gz) and extract the archive.

Create a symbolic link to the prank script, e.g:

ln -s p2rank_2.5/prank /usr/bin/prank

Modify the prank script to work with a symbolic link. At line 22, replace:

THIS_SCRIPT_DIR_REL_PATH=`dirname "${BASH_SOURCE[0]}"`

by

THIS_SCRIPT_DIR_REL_PATH=$(dirname "$(readlink -f "${BASH_SOURCE[0]}")")

Now, ASMC can use P2RANK to detect ligand binding pockets.

Docker Image

Download the Docker image in the latest GitHub release: https://github.com/labgem/ASMC/releases

Then, use the following command to load an image from the archive:

docker load -i asmc-latest.tar

Quick Usage

Run ASMC in a blind way (unknown active site) using a multi fasta file that should contain at least 100 sequences for clustering to be sufficiently relevant.

asmc run --log run_asmc.log --threads 6 -r reference_file -s sequences.fasta

reference_file should contains the path to the reference(s) structure(s), e.g:

<path>/RefA.pdb
<path>/RefB.pdb

NB: For more details, see the wiki