rinikerlab / lightweight-registration

LWReg: a lightweight chemical registration system

This provides a basic registration system which can be used either as a python library or via a command-line interface.

Basic operations:

• initdb: resets the database. Note that this destroys any information which is already in the database, so be careful with it.
• register: standardize the input molecule, calculates a hash for it, and adds them molecule to the database if it's not already there. returns the molregno (registry ID) of the newly registered molecule.
• query: takes a molecule as input and checks whether or not a matching molecule is registered. returns molregnos (registry IDs) of the matching molecule(s), if any
• retrieve: takes one or more IDs and returns the registered structures for them

Quick-start installation for non-experts:

Assuming that you have conda (or mamba or something equivalent) installed you can install lwreg directly from this github repo by first creating a conda environment with all the dependencies installed:

% conda env create --name py311_lwreg --file=https://raw.githubusercontent.com/rinikerlab/lightweight-registration/main/environment.yml

If you have mamba installed, you can run this instead (it will run faster):

% mamba env create --name py311_lwreg --file=https://raw.githubusercontent.com/rinikerlab/lightweight-registration/main/environment.yml

You can then activate the new environment and install lwreg:

% conda activate py311_lwreg
% python -m pip install git+https://github.com/rinikerlab/lightweight-registration

You can then verify that the install worked by doing:

% lwreg --help

If you want to use PostgreSQL as the database for lwreg, then you will also need to install the python connector for PostgreSQL:

% conda install -c conda-forge psycopg2

Installation for non-experts:

Please look at the INSTALL.md file.

Installation for experts:

Dependencies:

• rdkit v2023.03.1 or later
• click
• psycopg2 (only if you want to use a postgresql database)

After installing the dependencies (above) and checking out this repo, run this command in this directory:

pip install --editable .

Run in Docker

docker build -t lwreg .

# Run Jupyter notebook on the docker container
docker run -i -t -p 8888:8888 rdkit-lwreg /bin/bash -c "\
    apt update && apt install libtiff5 -y && \
    pip install notebook && \
    jupyter notebook \
    --notebook-dir=/lw-reg --ip='*' --port=8888 \
    --no-browser --allow-root"

Very basic usage demo

Command line

% lwreg initdb --confirm=yes
% lwreg register --smiles CCOCC
1
% lwreg register --smiles CCOCCC
2
% lwreg register --smiles CCNCCC
3
% lwreg register --smiles CCOCCC
ERROR:root:Compound already registered
% lwreg query --smiles CCOCCC
2
% lwreg retrieve --id 2
(2, '\n     RDKit          2D\n\n  0  0  0  0  0  0  0  0  0  0999 V3000\nM  V30 BEGIN CTAB\nM  V30 COUNTS 6 5 0 0 0\nM  V30 BEGIN ATOM\nM  V30 1 C 0.000000 0.000000 0.000000 0\nM  V30 2 C 1.299038 0.750000 0.000000 0\nM  V30 3 O 2.598076 -0.000000 0.000000 0\nM  V30 4 C 3.897114 0.750000 0.000000 0\nM  V30 5 C 5.196152 -0.000000 0.000000 0\nM  V30 6 C 6.495191 0.750000 0.000000 0\nM  V30 END ATOM\nM  V30 BEGIN BOND\nM  V30 1 1 1 2\nM  V30 2 1 2 3\nM  V30 3 1 3 4\nM  V30 4 1 4 5\nM  V30 5 1 5 6\nM  V30 END BOND\nM  V30 END CTAB\nM  END\n', 'mol')

Python

>>> import lwreg

>>> from lwreg import utils

>>> lwreg.set_default_config(utils.defaultConfig())   # you generally will want to provide more information about the database

>>> lwreg.initdb()
This will destroy any existing information in the registration database.
  are you sure? [yes/no]: yes
True

>>> lwreg.register(smiles='CCO')
1

>>> lwreg.register(smiles='CCOC')
2

>>> from rdkit import Chem

>>> m = Chem.MolFromSmiles('CCOCC')

>>> lwreg.register(mol=m)
3

>>> lwreg.register(mol=m)
---------------------------------------------------------------------------
IntegrityError                            Traceback (most recent call last)
Input In [10], in <cell line: 1>()
----> 1 lwreg.register(mol=m)
  
  ... DETAILS REMOVED ...

IntegrityError: UNIQUE constraint failed: hashes.fullhash

>>> lwreg.query(smiles='CCOC')
[2]

>>> lwreg.query(smiles='CCOCC')
[3]

>>> lwreg.query(smiles='CCOCO')
[]

>>> lwreg.retrieve(id=2)
{2: ('\n     RDKit          2D\n\n  0  0  0  0  0  0  0  0  0  0999 V3000\nM  V30 BEGIN CTAB\nM  V30 COUNTS 4 3 0 0 0\nM  V30 BEGIN ATOM\nM  V30 1 C 0.000000 0.000000 0.000000 0\nM  V30 2 C 1.299038 0.750000 0.000000 0\nM  V30 3 O 2.598076 -0.000000 0.000000 0\nM  V30 4 C 3.897114 0.750000 0.000000 0\nM  V30 END ATOM\nM  V30 BEGIN BOND\nM  V30 1 1 1 2\nM  V30 2 1 2 3\nM  V30 3 1 3 4\nM  V30 END BOND\nM  V30 END CTAB\nM  END\n',
  'mol')}

>>> lwreg.retrieve(ids=[2,3])
{2: ('\n     RDKit          2D\n\n  0  0  0  0  0  0  0  0  0  0999 V3000\nM  V30 BEGIN CTAB\nM  V30 COUNTS 4 3 0 0 0\nM  V30 BEGIN ATOM\nM  V30 1 C 0.000000 0.000000 0.000000 0\nM  V30 2 C 1.299038 0.750000 0.000000 0\nM  V30 3 O 2.598076 -0.000000 0.000000 0\nM  V30 4 C 3.897114 0.750000 0.000000 0\nM  V30 END ATOM\nM  V30 BEGIN BOND\nM  V30 1 1 1 2\nM  V30 2 1 2 3\nM  V30 3 1 3 4\nM  V30 END BOND\nM  V30 END CTAB\nM  END\n',
  'mol'),
 3: ('\n     RDKit          2D\n\n  0  0  0  0  0  0  0  0  0  0999 V3000\nM  V30 BEGIN CTAB\nM  V30 COUNTS 5 4 0 0 0\nM  V30 BEGIN ATOM\nM  V30 1 C 0.000000 0.000000 0.000000 0\nM  V30 2 C 1.299038 0.750000 0.000000 0\nM  V30 3 O 2.598076 -0.000000 0.000000 0\nM  V30 4 C 3.897114 0.750000 0.000000 0\nM  V30 5 C 5.196152 -0.000000 0.000000 0\nM  V30 END ATOM\nM  V30 BEGIN BOND\nM  V30 1 1 1 2\nM  V30 2 1 2 3\nM  V30 3 1 3 4\nM  V30 4 1 4 5\nM  V30 END BOND\nM  V30 END CTAB\nM  END\n',
  'mol')}

Custom standardization/filtering functions

When using the Python API you have extensive control over the standardization and validation operations which are performed on the molecule.

Start with a couple of examples showing what the 'fragment' and 'charge' built-in standardizers do:

>>> config['standardization'] = 'fragment'
>>> Chem.MolToSmiles(lwreg.standardize_mol(Chem.MolFromSmiles('CC[O-].[Na+]'),config=config))
'CC[O-]'
>>> config['standardization'] = 'charge'
>>> Chem.MolToSmiles(lwreg.standardize_mol(Chem.MolFromSmiles('CC[O-].[Na+]'),config=config))
'CCO'

Now define a custom filter which rejects (by returning None) molecules which have a net charge and then use that:

>>> def reject_charged_molecules(mol):
...     if Chem.GetFormalCharge(mol):
...         return None
...     return mol
...
>>> config['standardization'] = reject_charged_molecules
>>> Chem.MolToSmiles(lwreg.standardize_mol(Chem.MolFromSmiles('CC[O-].[Na+]'),config=config))
'CC[O-].[Na+]'

Here's an example which fails:

>>> lwreg.standardize_mol(Chem.MolFromSmiles('CC[O-]'),config=config) is None
True

We can chain standardization/filtering operations together by providing a list. The individual operations are run in order. Here's an example where we attempt to neutralise the molecule by finding the charge parent and then apply our reject_charged_molecules filter:

>>> config['standardization'] = ['charge',reject_charged_molecules]
>>> lwreg.standardize_mol(Chem.MolFromSmiles('CC[O-]'),config=config) is None
False
>>> lwreg.standardize_mol(Chem.MolFromSmiles('CC[N+](C)(C)C'),config=config) is None
True

That last one failed because the quarternary nitrogen can't be neutralized.

There are a collection of other standardizers/filters available in the module lwreg.standardization_lib

Registering conformers

When the configuration option registerConformers is set to True, lwreg expects that the compounds to be registered will have an associated conformer. The conformers are tracked in a different table than the molecule topologies and expectation is that every molecule registered will have a conformer (it's an error if they don't). It is possible to register multiple conformers for a single molecular structure (topology).

Note that once a database is created in registerConformers mode, it probably should always be used in that mode.

Differences when in registerConformers mode

• register() and bulk_register() require molecules to have associated conformers. Both return (molregno, conf_id) tuples instead of just molregnos
• query(): if called with the ids argument, this will return all of the conformers for the supplied molregnos as (molregno, conf_id) tuples. If called with a molecule, the conformer of the molecule will be hashed and looked up in the conformers`` table, returns a list of (molregno, conf_id)` tuples.
• retrieve(): if called with (molregno, conf_id) tuple(s), this will return a dictionary of (molblock, 'mol') tuples with (molregno, conf_id) tuples as keys where the molblocks contain the coordinates of the registered conformers.

Hashing conformers for registration

Just as molecular hashes are used to recognize when two molecules are the same, lwreg uses a hashing scheme to detect when two conformers are the same. The algorithm for this is simple: The atomic positions are converted into strings (rounding the floating point values to a fixed, but configurable, number of digits), sorting the positions, and then combining them into a single string. A SHA256 hash of this string is generated to give the final conformer hash.

If registering a multi-conformer molecule, it is most efficient to call register_multiple_conformers(). That only does the work of standardizing the molecule and calculating the molecule hash once.

Data layout

The base tables

Here's the SQL to create the base lwreg tables in sqlite:

create table registration_metadata (key text, value text);
create table hashes (molregno integer primary key, fullhash text unique, 
            formula text, canonical_smiles text, no_stereo_smiles text, 
            tautomer_hash text, no_stereo_tautomer_hash text, "escape" text, sgroup_data text, rdkitVersion text);
create table orig_data (molregno integer primary key, data text, datatype text, timestamp DATETIME DEFAULT CURRENT_TIMESTAMP);
create table molblocks (molregno integer primary key, molblock text, standardization text);

The conformers table

Here's the SQL to create the conformers table in sqlite when registerConformers is set:

create table conformers (conf_id integer primary key, molregno integer not null, 
                   conformer_hash text not null unique, molblock text);