ORIGINAL RESEARCH article
Front. Cell. Infect. Microbiol.
Sec. Clinical and Diagnostic Microbiology and Immunology
Volume 15 - 2025 | doi: 10.3389/fcimb.2025.1641385
Whole-genome Sequencing and Machine Learning Reveal Key Drivers of Delayed Sputum Conversion in Rifampicin-Resistant Tuberculosis
Provisionally accepted
- 1Cosmos Wisdom Mass Spectrometry Center of Zhejiang University Medical School, Hangzhou, China
- 2The First Affiliated Hospital of Ningbo University, Ningbo, China
- 3Fountain Valley High, Fountain Valley, United States
- 4Ningbo Municipal Center for Disease Control and Prevention, Ningbo, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Rifampicin-resistant tuberculosis (RR-TB) remains a major global health challenge, with delayed sputum culture conversion (SCC) predicting poor treatment outcomes. This study integrated whole-genome sequencing (WGS) and machine learning to identify clinical and genomic determinants of SCC failure in 150 RR-TB patients (2019-2023). Phenotypic and genotypic analysis revealed high rates of isoniazid resistance (74.0%) and rpoB mutations (97.3%, predominantly Ser450Leu), with 90% of strains belonging to Lineage 2 (Beijing family). While 64.7% achieved 2-month SCC, 18.0% remained culture-positive at 6 months. Univariate analysis linked 2-month SCC failure to smear positivity, resistance to isoniazid, amikacin, capreomycin, and levofloxacin, and pre-XDR-TB status, though only smear positivity (aOR=2.41, P=0.008) and levofloxacin resistance (aOR=2.83, P=0.003) persisted as independent predictors in multivariable analysis. A Random Forest model achieved robust prediction of SCC failure (AUC: 0.86±0.06 at 2 months; 0.76±0.10 at 6 months), identifying levofloxacin resistance (feature importance: 6.37), embB_p.Met306Ile (5.94), and smear positivity (5.12) as top 2-month predictors, while katG_p.Ser315Thr (4.85) and gyrA_p.Asp94Gly (3.43) dominated 6-month predictions. These findings underscore smear positivity, levofloxacin resistance, and specific resistance mutations as critical drivers of SCC failure, guiding targeted RR-TB treatment strategies.
Keywords: Rifampicin-resistant tuberculosis, Whole-genome sequencing, sputum culture conversion, machine learning, Drug resistance mutations
Received: 05 Jun 2025; Accepted: 22 Jul 2025.
Copyright: © 2025 Li, Fang, Lu, Gao, Wu, Chen and Che. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Yang Che, Ningbo Municipal Center for Disease Control and Prevention, Ningbo, China
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.