Abstract
Tuberculosis (TB) is still a significant public health threat, with rising drug resistance and high incidence in multiple areas worldwide. In the search for novel antitubercular agents, this study explores the application of a bioactivity-guided molecular networking approach to identify bioactive compounds from seven plant species (Curatella americana, Davilla nitida, Dipteryx punctata, Indigofera suffruticosa, Quassia amara, Tetradenia riparia, and Zingiber zerumbet) collected in French Guiana. Using ultrasound-assisted extraction followed by liquid-liquid partitioning and UHPLC-HRMS/MS analysis, a library of 72 samples was tested against Mycobacterium tuberculosis. The non-polar fractions from Indigofera suffruticosa, Tetradenia riparia, and Zingiber zerumbet showed the highest activity. The integration of metabolomic and bioassay data on molecular networks allowed the prioritization and annotation of active compounds, revealing flavonoids as contributors to the antitubercular activity of the active samples. In addition, the use of computational tools such as GNPS, SIRIUS, and TIMA-R enabled dereplication and increased the confidence in the structural prediction of active metabolites. This approach demonstrated its potential in accelerating the identification of both known and novel bioactive compounds without requiring exhaustive isolation, offering a robust strategy for natural product-based drug development against TB.