molspotter 0.0.1.dev1

Identify silly molecules

molspotter

Disclaimer

This repository is based on the awesome silly_walks work of Patrick Walters.

Summary

Identify silly molecules.

molspotter helps finding unusual/silly atom arrangements in molecules. To do so a reference dataset of molecules is used. Any fingerprint bit of a new molecule not appearing in the reference set is considered silly.

Installation

pip install molspotter

Usage

Three prepared spotters are included in molspotter. They are respectively trained on:

• ChEMBL (version 32)
• ExCAPE-DB (Zenodo)
• Papyrus (version 05.6)

API

One can easily load a spotter as follows:

from molspotter import SillyMolSpotter

sms = SillyMolSpotter.from_pretrained('chembl')

New molecules can easily be scored using this spotter instance:

from rdkit import Chem

mol = Chem.MolFromSmiles('C=C=C1CN(C)CCN1Cc5ccc(C(=O)Nc4ccc(C)c(Nc3nccc(c2cccnc2)n3)c4)cc5')
sms.score_mol(mol)
# 0.06666666666666667

The silly bit score is defined as $score = \frac{n_{silly bits}}{n_{on bits}}$. A value of 0.0 means that the molecule does not contain any silly bit, while a value of 1.0 means the molecule is only made of silly bits.

Silly bits can also be displayed:

img = sms.show_mol(mol)

One can create their own spotter using the following:

# Instantiate a spotter with radius of Morgan fingerprint value of 2
sms = SillyMolSpotter(fp_radius=2)

sms.add_file('PATH_TO_MOLECULAR_FILE')

Finally, a spotter can be saved and loaded back from disk:

sms.save('PATH_TO_SAVE_SPOTTER_TO')

sms2 = SillyMolSpotter.from_file('PATH_TO_SAVED_SPOTTER')

Command-line interface

• Create a spotter

molspotter create -o examples/chembl_drugs.sp -i examples/chembl_drugs.sd

• Score with a spotter

molspotter score -s examples/chembl_drugs.sp -i examples/chembl_drugs.sd
molspotter score -s chembl -i examples/chembl_drugs.sd