High throughput screening (HTS) for compounds with activity against Mycobacterium tuberculosis whole cells is being pursued widely as a step towards new tuberculosis (TB) drugs. Hit rates from these screening campaigns typically range from less than 1% to low single digits. New approaches to increase the efficiency of hit identification and prioritization are urgently needed. We have recently built and validated Bayesian machine learning models for predicting compounds with activity (and counter-screens predicting vero cell toxicity) against M. tuberculosis based on publicly available large scale high throughput screening data from TAACF.  We now demonstrate prospective validation across independent laboratories in which after computational screening of 82,403 molecules, a total of 550 molecules were tested in vitro and 124 actives were identified. We describe hit rates for the datasets from 15-28%. These models can be used to increase screening efficiency by any group pursuing broad screening for M. tuberculosis drug discovery.

Enhancing high throughput screening for <i>Mycobacterium tuberculosis</i> drug discovery using Bayesian models

245th National Meeting (2013)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.

Enhancing high throughput screening for Mycobacterium tuberculosis drug discovery using Bayesian models