RDKit Performance on Substructure Search

Cheminformatics Tutorials - Herong's Tutorial Examples

∟RDKit Performance on Substructure Search

This chapter provides introductions and tutorial examples on substructure search performance with RDKit. Topics include introduction of large molecule dataset; substructure search on one molecule at a time taking a long time; substructure search on with a molecule object library taking a large memory; substructure search by reading molecule SMILES library performing the best.

These sections are omitted from this Web preview version. To view the full content, see information on how to obtain the full version this book.

Large Molecule Datasets for Performance Testing

CPU Info and Usage for Performance Testing

Molecule Construction Performance with RDKit

Substructure Search Performance with RDKit

Substructure Library Search Performance with RDKit

Binary Library Search Performance with RDKit

SMILES Library Search Performance with RDKit

Trusted SMILES Library Search Performance with RDKit

Library File for Substructure Search with RDKit

Takeaways:

chembl26_very_active.sdf.gz is large dataset with 35767 molecules retrieved from ChEMBL database.
A computer with 3.688 GHz CPU and 8 GB RAM is used to do performance tests.
Doing substructure search with one molecule at a time requires 4 minutes and 63 MB memory for 1,000,000 molecules.
Doing substructure search with a molecule library in binary format requires 4 minutes and 674 MB memory for 1,000,000 molecules.
Doing substructure search with a molecule library in SMILES (or trusted SMILES) format requires 5 minutes and 170 MB memory for 1,000,000 molecules.
Doing substructure search by reading a molecule library in trusted SMILES format requires 10 seconds and 118 MB memory for 1,000,000 molecules.