The TB27 Transcriptomic Model for Predicting Mycobacterium tuberculosis Culture Conversion.

Area of Science:

• Microbiology
• Genomics
• Machine Learning

Background:

• Monitoring tuberculosis (TB) treatment is challenging due to the slow growth of Mycobacterium tuberculosis.
• Host RNA signatures offer a promising approach for tracking treatment response in TB patients.

Purpose of the Study:

• To identify and validate a whole blood-based RNA signature for predicting microbiological treatment responses in tuberculosis patients.
• To develop a machine learning algorithm for predicting time to culture conversion during anti-tuberculosis therapy.

Main Methods:

• A multi-step machine learning algorithm was employed to identify an RNA signature.
• The algorithm was developed using a 149-patient training and testing cohort, resulting in a 27-gene signature (TB27).
• External validation was performed on a separate cohort of 34 patients.

Main Results:

• The TB27 signature demonstrated high accuracy in predicting the time to culture conversion (TCC).
• In the test dataset, predicted TCC and observed TCC showed a correlation coefficient of r=0.98.
• The external validation cohort also showed a strong correlation (r=0.98) between predicted and observed TCC.

Conclusions:

• A validated whole blood-based RNA signature (TB27) shows excellent agreement for predicting Mycobacterium tuberculosis culture conversion times.
• TB27 is a potential biomarker for advancing anti-tuberculosis drug development.
• This signature may improve the prediction of treatment responses in clinical practice.