What is it about?
We used mycobacterium tuberculosis whole cell activity and cytotoxicity data to build bayesian models. These were then used to screen a vendor library, select and test compounds. We also filtered a library of GSK antimalarial hits and identified 5 compounds active against Mtb out of 7 tested. One triazine compound tested had an MIC of 62.5 ug/ml. The compound was not active in vivo but had good kill kinetics.
Why is it important?
We demonstrate that we can identify whole cell active and not cytotoxic molecules against Mtb using machine learning. We can leverage the public data to select new compounds that are active in vitro. This study unearthed an active series which had been first published 40 years earlier and proposed its utility.
The following have contributed to this page: Dr Sean Ekins and Joel Freundlich