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.
The clustering step can be carried out on either structural or sequence alignment and users can directly analyse their own set of protein 3D structures (e.g. AI-based models) by skipping the homology modeling step.
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.
- 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)
Download the p2rank tar.gz file (e.g: p2rank_2.5.tar.gz) and extract the archive.
Create a symbolic link related to the prank script, e.g:
ln -s <full_path_to>/p2rank_2.5/prank /usr/bin/prank
Modify the previous 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.
Follow the instructions in the Docker section of the wiki
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