As of May 2021, CMILES is no longer being actively maintained. Most of its essential functionality has been migrated to the Open Force Field Toolkit and QCSubmit packages.
API points that handle former CMILES functions in these packages include:
mapped=True
keyword argumentGenerate canonical identifiers for chemical databases, specifically quantum chemical data.
cmiles seeks to address several issues:
Link the QC molecule to its cheminformatics molecular graph to make QC data
useful to the force field and machine learning communities.
A QC molecule is identified by its elements and geometry. Each conformer
is considered a separate molecule. However, from a cheminformatics perspective,
the connectivity graph is the identity of the molecule.
Canonical identifiers.
When indexing a chemical database, it is crucial that the identifier
is canonical to reduce redundancy and search failures.
Problem:
Each toolkit has its own canonicalization algorithm. This algorithm may
change in different versions of the toolkit.
Solution:
Distribute cmiles as a Docker image with pinned dependencies.
Canonical order of QC geometry.
Since QC molecules are identified by their elements and geometry, canonical
atomic order can reduce redundancy.
Problem:
In cheminformatics toolkits, the order of atoms in a molecular graph are arbitrary.
Therefore, every time you generate a new molecular graph, the order of the atoms can change.
Solution:
Use the toolkits canonical atomic ranking to generate a SMILES string that has
explicit hydrogens and map indices that correspond to the canonical atomic ranking.
Then, this mapped SMILES can be used as a SMARTS pattern to find the mapping of
a molecular graph's these map indices to atomic indices. This map is then used
to generate canonical order for the xyz coordinates and symbols in the QC molecule.
cmiles.utils
provides functions that will do this.
In addition, the mapped SMILES ensure that all molecular graphs generated can be mapped
to QC geometries if those geometries are in the order of the map.
Standardize compounds
Provide an index to find all protomers of a molecule.
Problem:
Different protomeric states are different QC molecules. However, sometimes all
protomers and / tautomers of a molecule are needed. SMILES strings are different
for each protomer. While InChI standardizes molecular charge states, it only
standardize the tautomers it recognizes and not others. Some tautomers
it does not capture are keto-enol and enamine-imine.
cmiles
does not yet offer a full solution. Currently it provides the InChI, the
unique protomer from openeye
and MolStandardize
from rdkit. rdkit's solution only addresses tautomers of the
same charge states and openeye's solution does not capture indoles, isoindoles and
some mesomers. However, the union of these identifiers captures more than each individual solution.
cmiles
has one main function, to_molecule_id(molecule)
This
function returns a dictionary with 10 identifiers generated with either OpenEye
or RDKit.
cmiles.utils
provides functions to generate atom maps and map ordered geometries.
One of the following cheminformatics toolkits:
Copyright (c) 2018, Chaya D. Stern
Project based on the Computational Chemistry Python Cookiecutter