Combining Metabolite-Based Pharmacophores with Bayesian Machine Learning Models for Mycobacterium tuberculosis Drug Discovery

  • Sean Ekins, Peter B. Madrid, Malabika Sarker, Shao-Gang Li, Nisha Mittal, Pradeep Kumar, Xin Wang, Thomas P. Stratton, Matthew Zimmerman, Carolyn Talcott, Pauline Bourbon, Mike Travers, Maneesh Yadav, Joel S. Freundlich
  • PLoS ONE, October 2015, Public Library of Science (PLoS)
  • DOI: 10.1371/journal.pone.0141076

Mtb pharmacophores and drug discovery

What is it about?

After some bioinformatics to identify potential targets, we used pharmacophores for over 60 metabolites and substrates in Mtb to search a vendor library, these compounds were scored with Bayesian models, ~100 were tested and a quinoxaline di N-oxide was used as a starting point for medicinal chemistry optimization. We generated some ADME data and one compound was tested in vivo and had poor PK data.

Why is it important?

We used a combined bioinformatics and cheminformatics approach alongside medicinal chemistry and biology screening. One compound tested was also active against drug resistant strains of Mtb. This provides additional data for pursuit of quinoxaline di N-oxide as a novel class of Mtb inhibitors. The key challenge is resolving ADME/PK.

Perspectives

Dr Sean Ekins
Collaborations in Chemistry

This was a collaborative project which merged several different groups and techniques for Mtb drug discovery. It further shows how computational data can be useful for identifying compounds of interest. This work represents a phase II STTR and lead to the development of an API for the CDD Vault software as well as additional data in the public domain on compounds and targets for Mtb.

Read Publication

http://dx.doi.org/10.1371/journal.pone.0141076

The following have contributed to this page: Dr Sean Ekins