New insights toward the discovery of antibacterial agents: multi-tasking QSBER model for the simultaneous prediction of anti-tuberculosis activity and toxicological profiles of drugs.

Tuberculosis (TB) constitutes one of the most dangerous and serious health problems around the world. It is a very lethal disease caused by microorganisms of the genus mycobacterium, principally Mycobacterium tuberculosis (MTB) which affects humans. A very active field for the search of more efficient anti-TB chemotherapies is the use in silico methodologies for the discovery of potent anti-TB agents. The battle against MTB by using antimicrobial chemotherapies will depend on the design of new chemicals with high anti-TB activity and low toxicity as possible. Multi-target methodologies focused on quantitative-structure activity relationships (mt-QSAR) have played a very important role for the rationalization of drug design, providing a better understanding about the molecular patterns related with diverse pharmacological profiles including antimicrobial activity. Nowadays, almost all mt-QSAR models have considered the study of biological activity or toxicity separately. In the present study, we develop by the first time, a unified multitasking model based on quantitative-structure biological effect relationships (mtk-QSBER) for the simultaneous prediction of anti-TB activity and toxicity against Mus musculus and Rattus norvegicus. The mtk-QSBER model was created by using linear discriminant analysis (LDA) for the classification of compounds as positive (high biological activity and/or low toxicity) or negative (otherwise) under many experimental conditions. Our mtk-QSBER model, correctly classified more than 90% of the case in the whole database (more than 12,000 cases), serving as a powerful tool for the computer-assisted screening of potent and safe anti-TB drugs.