Introduction

Molcomplib is a key component of the MolCompass project. It is a python library, that provides the pretrained parametric t-SNE model for molecular visualization. This library generates X and Y coordinates for compounds so that similar compounds group together forming well-recognisible clusters. The library is lightweight, and requires only rdkit and numpy (optionally, pandas)

Graphical abstract

Application of a Parametric t-SNE model: A set of chemical compounds (A) is converted into ECFP binary fingerprints of a fixed length (B). Then, a pretrained artificial neural network (ANN) (C) projects these fingerprints into coordinates, forming 2D clusters where structurally similar compounds are grouped together (D)

Installation

pip install molcomplib

Usage

Import the libraty and create MolCompass object:

from molcomplib import MolCompass
compass = MolCompass()

You can apply it for an individual molecule:

print(compass('CCO'))

[ -0.98872091 -35.78184447]

Or apply to a list of molecules:

list_of_compounds = ['CCC','CCO','C1=CC=C(C=C1)C=O']
res = np.vstack([compass(compound) for compound in list_of_compounds])
print(res)

[[ -1.60652074 -36.25469236]
 [ -0.98872091 -35.78184447]
 [-32.6078482   -7.50528324]]

molcomplib also provides an interface to pandas interface:

from molcomplib import MolCompass
molcomp = MolCompass()
import pandas as pd
df = pd.DataFrame({'smiles':["CCO","CCC"]})
res = molcomp.process(df)
print(res)

        x          y
  smiles         x          y
0    CCO -0.988721 -35.781844
1    CCC -1.606521 -36.254692

Other parts of MolCompass

molcomplib is a computational engine for MolCompass KNIME Node and MolCompass Viewer GUI tool (molcompview).

Citation

Sosnin, S. MolCompass: multi-tool for the navigation in chemical space and visual validation of QSAR/QSPR models. J Cheminform 16, 98 (2024). https://doi.org/10.1186/s13321-024-00888-z

Citation (bibtex)

@article{Sosnin2024,
  title = {MolCompass: multi-tool for the navigation in chemical space and visual validation of QSAR/QSPR models},
  volume = {16},
  ISSN = {1758-2946},
  url = {http://dx.doi.org/10.1186/s13321-024-00888-z},
  DOI = {10.1186/s13321-024-00888-z},
  number = {1},
  journal = {Journal of Cheminformatics},
  publisher = {Springer Science and Business Media LLC},
  author = {Sosnin,  Sergey},
  year = {2024},
  month = aug 
}

DOI for the package version