Predicting drug interactions from chemogenomics using INDIGO

Abstract

Designing effective antibiotic combination regimens is critical for countering drug resistance in pathogens. Yet the large combinatorial search-space makes the identification of effective combinations a significant challenge. There is a great need for computational approaches that can rapidly prioritize potential combination regimens based on the antagonistic and synergistic interactions among the constituent antibiotics. This protocol outlines the steps to predict antibiotic interactions from chemogenomics data using the INDIGO algorithm. INDIGO predicted novel drug–drug interaction outcomes quantitatively with high accuracy based on experimental evaluation of predictions in E. coli and S. aureus, and it overcomes several limitations of existing drug-interaction prediction algorithms. The INDIGO approach also expands the applicability of chemogenomic data from model organisms to a broader set of less-studied pathogens. INDIGO can predict drug-interaction outcomes in the bacterial pathogens S. aureus and M. tuberculosis, using chemogenomics data from E. coli by quantifying the degree of conservation of the drug–gene interaction network between different species. The INDIGO approach, which is demonstrated for E. coli and S. aureus in this protocol, can be applied easily to other organisms including pathogens.