The Next Era of Pharmaceutical Research: From Bayesian Models to Deep Learning

Over the past decade our academic collaborations have encompassed applying machine learning to drug discovery for neglected diseases like Ebola, Chagas disease and tuberculosis. These have led to the discovery of drug-like molecules and repurposed drugs with in vivo activity identified using Bayesian machine learning methods. In turn these efforts have resulted in grant funding from the NIH to further explore the molecules discovered. Additional recent academic collaborations include OpenZika, which has fostered a dynamic and fruitful interaction with our colleagues in Brazil and the support of IBM World Community Grid. These efforts promise to develop new molecules for neglected diseases. More recently we have been exploring a wider array of machine learning approaches and this has coincided with the increase in use of deep neural networks (DNN). We have explored extensive curation of public datasets for neglected tropical diseases and compared many different algorithms and descriptors.