Loading...
Discovery of novel InhA inhibitors through structural bioinformatics and machine learning-driven QSAR screening of natural products
Anthony, Godswill Imolele Nwosu, Samuel Nzube Owolade, Adedoyin John-Joy Kehinde, Temitope Olubanjo Teye, Richard Gamah Akor, Samuel Eneojo Fabuyi, Favour Segun Nwafor, Ifeoma Roseline Oleh, Precious Kelechi Ilori, Toluwalope Daniel
Anthony, Godswill Imolele
Nwosu, Samuel Nzube
Owolade, Adedoyin John-Joy
Kehinde, Temitope Olubanjo
Teye, Richard Gamah
Akor, Samuel Eneojo
Fabuyi, Favour Segun
Nwafor, Ifeoma Roseline
Oleh, Precious Kelechi
Ilori, Toluwalope Daniel
Advisors
Editors
Other Contributors
EPub Date
Publication Date
2026-03-30
Submitted Date
Collections
Files
Loading...
Published version
Adobe PDF, 2.77 MB
Other Titles
Abstract
Tuberculosis (TB) persists as a formidable global health challenge, particularly due to the emergence of multidrug-resistant (MDR) and extensively drug-resistant (XDR) variants that limit the effectiveness of existing therapies. These variants limit therapeutic options, prolong treatment duration and increase the risk of treatment failure. Additionally, the antitubercular drug discovery remains relatively limited; moreover, the rise in drug-resistant strains necessitates the need to identify and develop novel drug candidates that can overcome these resistance patterns. This study aims to design novel InhA inhibitors that can bind to and inhibit InhA and circumvent the resistance pathway in the KatG enzyme of Mycobacterium tuberculosis. We sourced an extensive library of 276 518 natural products from the LOTUS database and screened the compounds through a series of rigorous computational approaches such as drug-likeness filtration, molecular docking, MM-GBSA analysis, and ADMET prediction. We developed a machine learning model using a Message Passing Neural Network (MPNN). The MPNN was trained to predict bioactivity profiles of 31 597 natural products against the InhA enzyme. Molecular dynamics simulations further confirmed the stability of these interactions over 100-nanoseconds. Out of the screened compounds, four novel drug candidates exhibited strong binding affinities with binding energies of −12.204 kcal/mol, −11.926 kcal/mol, −11.624 kcal/mol, and −11.548 kcal/mol, respectively, surpassing the co-crystallized ligand (−8.895 kcal/mol), and the standard drug, Isoniazid (−12.204 kcal/mol). The top hit molecules demonstrated high and considerable structural stability during molecular dynamics simulations. Additionally, the pharmacokinetic profile of LTS0161715 exhibited low toxicity. positioning LTS0161715 for further investigation. These research findings elucidate the potential of direct InhA inhibitors, particularly LTS0161715, as a promising drug candidate for anti-tubercular drug development, while also highlighting the need for further optimization to enhance safety and efficacy.
Citation
Anthony, G. I., Nwosu, S. N., Owolade, A. J.-J., Kehinde, T. O., Teye, R. G., Akor, S. E., Fabuyi, F. S., Nwafor, I. R., Oleh, P. K., Ilori, T. D., & Bodun, D. S. (2026). Discovery of novel InhA inhibitors through structural bioinformatics and machine learning-driven QSAR screening of natural products. Biology Methods and Protocols, 11(1), bpag015. https://doi.org/10.1093/biomethods/bpag015
Publisher
Oxford University Press
Journal
Biology Methods and Protocols
Research Unit
PubMed ID
PubMed Central ID
Type
Article
Language
Description
© The Author(s) 2026. Published by Oxford University Press.
Series/Report no.
ISSN
EISSN
2396-8923