EnzyMM - The Enzyme Motif Miner

<!-- [](https://badge.dimensions.ai/details/id/pub.1147419140) --> <!-- [](https://aur.archlinux.org/packages/python-pyhmmer) -->

📚 Full documentation is availabe here: https://enzymm.readthedocs.io/en/latest/

️Overview

Enzyme Motif Miner uses geometric template matching to identify known arrangements of catalytic residues called templates in protein structures. It searches protein structures provided by the user against a database of templates. EnzyMM ships with a library of catalytic templates derived from the Mechanism and Catalytic Site Atlas (M-CSA) but you can also generate your own. These templates represent consensus arrangements of catalytic sites found in active sites of experimental protein structures.

As catalytic sites are both highly conserved and absolutely critical for the function of a protein, identifying them offers many biological insights. This method has two key advantages. Firstly, as it doesn't rely on sequence or (global) fold similarity, similar catalytic arrangements can be found accross great evolutionary distances offering insights into the divergence or even convergence of enyzmes. Secondly, as geometric matching is very fast, EnzyMM scales along side databases of predicted protein structures. Expect to scan a protein structure in a matter of seconds on consumer laptops.

As a database driven method, EnzyMM is inherently limited by the coverage of residue arrangements in its template library. The provided template library covers nearly the entire M-CSA and thus around 3/4 of enzyme mechanisms classified by the Enzyme Commission to the 3rd level. Catalytic arrangements not found in the PDBe won't be included in the M-CSA. Of course, the user can also provide their own library of templates. While primarily intended for catalytic sites, you are invited to search with your own library of templates.

For the actual geometric matching EnzyMM relies on PyJess - a Cython wrapper of Jess.

If you just want to try EnzyMM we provide a webserver at https://www.ebi.ac.uk/thornton-srv/m-csa/enzymm .

🔧 Installing EnzyMM

EnzyMM is implemented in Python, and supports all versions from Python 3.7 on Linux and MacOS. It requires additional libraries that can be installed directly from PyPI, the Python Package Index.

Use pip to install EnzyMM on your machine:

$ pip install enzymm

Alternatively, use Anaconda to install EnzyMM. Optionally install gemmi to search CIF files too:

$ conda install -c bioconda enzymm gemmi

This will both install EnzyMM and also download a library of catalytic templates together with important metadata. This requires around 16MB of data to be downloaded. It should also run on windows (though this is not tested for on release).

🖼️ Images

Lightweight images built from python:3.13-alpine are available:

Pull the latest Docker image from GHCR:

docker pull ghcr.io/rayhackett/enzymm:latest

Pull the latest Apptainer image via ORAS from GHCR:

apptainer pull oras://ghcr.io/rayhackett/enzymm:latest

🔎 Running EnzyMM

Once EnzyMM is installed, you can run it from the terminal. The user can either provide a path to a single protein structure -i or to run multiple queries at once, the path to a text file -l which itself contains a list of paths to protein structures. Structures are accepted in both CIF/mmCIF and PDB file format. Optionally, an output directory for PDB structures of the identified matches per query protein can be supplied with the --pdbs flag.

$ enzymm -i some_structure.pdb -o results.tsv --pdbs dir_to_save_matches

Additional parameters of interest are:

• --per-residue-results, which will return an additional table mapping each matched residue to its original annotation and catalytic role.
• --jobs or -j, which controls the number of threads used to parallelize the search. By default, it will use one thread less than available on your system using os.cpu_count.
• --unfiltered or -u, which disables filtering of matches by RMSD and residue orientation. By default, filtering is enabled.
• --skip-smaller-hits, which skips searches with smaller templates on a query if a match to a larger template has already been found.
• --parameters or -p, which controls the RMSD threshold and pairwise distance threshold applied. By default sensible thresholds are selected. Refer to the Docs for details
• --template-dir or -t, though which the user may supply their own template library. By default, a library of catalytic templates derived from the M-CSA is loaded.
• --conservation-cutoff or -c, which can be set to exclude atoms with B-factors or pLDDT scores below this threshold from matching. This is not set by default.

Further, EnyzMM is designed with modularity in mind and comes with a fully usable internal API. Please refer to the Documentation for further reference.

🖹 Results

Tabular results

EnzyMM will create a single output file by default. You can chose between the default full results or a more simple style with fewer columns using the flag --simple-results:

• {output}.tsv: A .tsv file containing a summary of all results. One row is printed per match.

If you pass the --per-residue-results flag, EnzyMM will additionally create a table with one line per matched residue, mapping each residue to its original annotations.

• {output}.residues.tsv: A .tsv file containing one row per residue per match.

If you pass the --parquet flag, EnzyMM will write .parquet files instead of .tsv files. Note that this additionally requires the polars library.

Aligned Structures

For visual exploration of matches, you can optionally save an alignment of the template and the matched query residues to a PDB file which can be viewed with any molecular viewer. To do so, supply an output directory after the --pdbs flag for the .pdb files.

What will get written depends of in the --transform flag is set or not:

• {pdbs_dir}/{query_identifier}_matches.pdb: Default: One .pdb file per query with matched residues in the query written in the query reference frame.
• {pdbs_dir}/{template_pdb_identifier}_matches.pdb: Default: One .pdb file per template structure which matches any query written in the template reference frame. In short, --transform forces the output into the template reference frame. Therefore only matches from the same template structure can be aligned which is why we write one file per matched template structure!

Add additional information to each .pdb file with the following fla