用机器学习辅助的拉曼光谱快速预测结核分枝杆菌耐药性

Aug 4, 2026

Upload by Himbo

Tuberculosis (TB) remains a leading cause of infectious-disease death, and delays in diagnosis and antibiotic susceptibility testing (AST) contribute to inappropriate regimens and emerging drug resistance. The study presents a rapid, label-free workflow using Raman spectroscopy combined with machine learning to identify Mycobacterium tuberculosis strains and antibiotic-resistant mutants. Over 20,000 single-cell spectra are used to train classifiers for resistance to isoniazid, rifampicin, moxifloxacin, and amikacin, enabling high-accuracy resistance profiling without antibiotic co-incubation on dried samples and moderate accuracy on dried patient sputum. A low-cost portable Raman microscope supports field deployment in TB-endemic regions.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages22

Reading Time55 min

AccessFree to download

Introduction

Need for rapid, culture-free AST

Limitations of existing methods

Raman spectroscopy and machine learning approach

Dataset and model training

Performance on dried samples and patient sputum

Low-cost portable Raman microscope for field deployment

Predicting tuberculosis drug resistance with machine learning-assisted Raman spectroscopy Authors: Babatunde Ogunlade 1* ✝ , Loza F. Tadesse2,3,4* ✝ , Hongquan Li5* ✝ , Nhat Vu6, Niaz Banaei7, Amy K. Barczak4,8,9, Amr. A. E. Saleh 1, 10, Manu Prakash2 and Jennifer A. Dionne 1, 11✝ Affiliations: 1 Department of Materials Science and Engineering, Stanford University; Stanford, 94305, CA, USA. 2 Department of Bioengineering, Stanford University School of Medicine and School of Engineering; Stanford, 94305, CA, USA. 3 Department of Mechanical Engineering, Massachusetts Institute of Technology; Cambridge, 02142, MA, USA. 4 The Ragon Institute, Massachusetts General Hospital; Cambridge, 02139, MA, USA. 5 Department of Applied Physics, Stanford University; Stanford, 94305, CA, USA. 6 Pumpkinseed Technologies, Inc; Palo Alto, 94306, CA, USA. 7 Department of Pathology, Stanford University School of Medicine; Stanford, 94305, CA, USA. 8 Division of Infectious Diseases, Massachusetts General Hospital; Boston, 02114, MA, USA. 9 Department of Medicine, Harvard Medical School; Boston, 02115, MA, USA. 10Department of Engineering Mathematics and Physics, Cairo University; Giza, 12613, Egypt. 11 Department of Radiology, Molecular Imaging Program at Stanford (MIPS), Stanford University School of Medicine; Stanford, 94035, CA, USA. *Indicates equal contribution ✝ To whom correspondence should be addressed; [E-mail: ](E-mail: bogun@stanford.edu)[bogun@stanford.edu](E-mail: bogun@stanford.edu); [hqli@stanford.edu](hqli@stanford.edu) ; [lozat@mit.edu](lozat@mit.edu) ; [jdionne@stanford.edu](jdionne@stanford.edu). One Sentence Summary: Raman spectroscopy and machine learning can be used to rapidly identify and differentiate strains of antibiotic-resistant tuberculosis (TB) with high accuracy, on a low-cost platform suitable for deployment in TB-endemic regions. Abstract: Tuberculosis (TB) is the world's deadliest infectious disease, with 1.5 million annual deaths and half a million annual infections. Rapid TB diagnosis and antibiotic susceptibility testing (AST) are critical to improve patient treatment and to reduce the rise of new drug resistance. Here, we develop a rapid, label-free approach to identify Mycobacterium tuberculosis (Mtb) strains and antibiotic-resistant mutants. We collect over 20,000 single-cell Raman spectra from isogenic mycobacterial strains each resistant to one of the four mainstay anti-TB drugs (isoniazid, rifampicin, moxifloxacin and amikacin) and train a machine-learning model on these spectra. On dried TB samples, we achieve > 98% classification accuracy of the antibiotic resistance profile, without the need for antibiotic co-incubation; in dried patient sputum, we achieve average classification accuracies of ~ 79% . We also develop a low-cost, portable Raman microscope suitable for field-deployment of this method in TB-endemic regions. Main text: INTRODUCTION The discovery of antibiotics in the early 20th century marked a turning point in our defense against tuberculosis (TB)– one of the deadliest infectious diseases known to humans. At that time, TB had become curable and was considered to be on a path toward elimination. However, by the end of the 20th century, TB re-emerged as a leading cause of death globally, in part due to challenges in diagnosis and its evolved resistance to antibiotics. While growth in liquid culture is considered the gold standard for organism identification and determination of antibiotic susceptibility, the causative organism, Mycobacterium tuberculosis (Mtb) can take up to 40 days to culture. As antibiotic susceptibility testing requires pathogen growth, definitive determination of Mtb antibiotic resistance using traditional antibiotic susceptibility testing (AST) can take additional weeks. In the delay between diagnosis and AST results, many patients are treated with only partially active antibiotic regimens, giving rise to resistant strains. I

FAQs

该研究如何实现结核分枝杆菌耐药性的快速判定？

通过拉曼光谱采集样本的单细胞光谱，并用机器学习模型训练来识别结核分枝杆菌菌株及其耐药突变谱。该方法为标签-free流程，可在无需抗生素共培养的条件下完成分类。

研究训练数据的规模与来源是什么？

研究收集了超过2万条单细胞拉曼光谱，来自同系（isogenic）分枝杆菌菌株，这些菌株分别对四种主要抗结核药物（异烟肼、利福平、莫西沙星、阿米卡星）中的一种耐药。

该方法在样本检测中的准确度表现如何？

在干燥样本上实现了超过98%的抗药性谱分类准确度，并且不需要抗生素共培养；在干燥患者痰液中，平均分类准确度约为79%。

Aug 4, 2026

Upload by Himbo