The identification of novel chemical probes represents a significant hurdle in Science's current resource-limited setting. This fate is magnified for infectious diseases such as tuberculosis and malaria. Using public datasets for compounds screened against these diseases we can generate Bayesian machine learning models that can be harnessed to virtually screen commercial libraries and identify chemical probes. We have demonstrated experimentally confirmed actives at hit rates exceeding 16%. We report the first dual-event Bayesian models utilizing antitubercular whole-cell and mammalian cell cytotoxicity data as well as a novel approach to predict antimalarials. The most potent antitubercular hit exhibits in vitro activity and in vitro/in vivo safety profile of a chemical probe. Bayesian models afford significant economies in time, effort, and cost. We discuss how this strategy may be broadly applicable to probe discovery for a range of infectious diseases.

Bayesian models to identify chemical probes for the study of infectious diseases

245th National Meeting (2013)

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