A conformal Bayesian network for classification of Mycobacterium tuberculosis complex lineages

Area of Science:

• Genetics
• Bioinformatics
• Epidemiology

Background:

• Tuberculosis (TB), caused by Mycobacterium tuberculosis Complex (MTBC), is a major global health concern.
• Accurate identification of MTBC genetic lineages is crucial for effective TB control and tracking.
• Mycobacterial Interspersed Repetitive Unit (MIRU) and Spacer Oligonucleotide Typing (spoligotyping) are key DNA fingerprinting methods for MTBC analysis.

Purpose of the Study:

• To develop and validate a novel Conformal Bayesian Network (CBN) for classifying MTBC strains into six major genetic lineages.
• To create a method that can utilize various combinations of available genetic biomarkers (MIRU and spoligotyping) for lineage classification.
• To provide insights into the genetic diversity of MTBC through the analysis of biomarker signatures.

Main Methods:

• Development of a Conformal Bayesian Network (CBN) model.
• Training the CBN on large historical TB databases comprising diverse subsets of MTBC biomarkers.
• Validation of the CBN model on three large MTBC collections (over 34,737 isolates) genotyped with different combinations of spoligotyping and MIRU (12 and 24 loci) data.

Main Results:

• The CBN model accurately classifies MTBC strains into major genetic lineages.
• The model demonstrates high performance across different combinations of available biomarkers, accommodating incomplete genotyping data.
• Distinct MIRU and spoligotype signatures associated with each lineage were identified, explaining the model's high accuracy and providing insights into MTBC genetic diversity.

Conclusions:

• The developed CBN effectively classifies MTBC lineages using available PCR-based biomarkers.
• The CBN method offers high performance and can be extended with new biomarkers for future TB tracking and control efforts.
• An online tool is available to facilitate the use of the CBN model in TB research and control programs.