The rapid spread of multi-drug resistance has created a great need for new combination therapies to treat a variety of conditions, including infectious diseases and cancer. In one pressing example, multidrug resistant tuberculosis (TB) affects about 500,000 people each year and novel drug regimens are sorely needed. However, identifying new regimens has been daunting in part due to the inability to prioritize among a very large number of possible drug combinations. To address this need, we have generated an experimentally grounded, machine learning algorithm, INDIGO-MTB, which predicts the synergy or antagonism of TB drug combinations with high accuracy. Here we propose to adapt INDIGO-MTB into a multifactorial pipeline to dissect combinatorial drug efficacy and drive preclinical regimen development for TB. We will build in and validate the ability to predict drug interactions under stressful environmental conditions that mimic TB infection, and extract molecular mechanisms of drug interactions. We will then combine synergy and efficacy measurements to create new regimen rankings, which we will validate both in vitro and in a mouse model of TB infection. Altogether, our work will establish a tool for rapid assessment of TB drug combinations and a framework for applying this approach to other conditions where new multidrug therapies are needed.
Tuberculosis is a massive public health problem, and new drug regimens are sorely needed. To address this need, we are assembling a multifactorial pipeline of experiments and computation to understand how drugs interact and to drive regimen development for TB. Our work will establish a tool for rapid assessment of TB drug combinations and a framework for applying this approach to other conditions where new multidrug therapies are needed.
