A comprehensive machine learning for high throughput Tuberculosis sequence analysis, functional annotation, and...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Tuberculosis (TB) is a major global health threat, causing over a million deaths annually.
• Accurate and early diagnosis of TB is crucial for effective treatment and disease control.
• Machine learning (ML) offers powerful tools for analyzing complex biological data, including RNA sequences.

Purpose of the Study:

• To develop and evaluate ML-supervised algorithms for classifying TB patients from RNA sequence count data.
• To identify significant genes and biological pathways associated with TB using feature importance techniques.
• To explore potential therapeutic targets and drug interactions through advanced computational analysis.

Main Methods:

• Trained multiple ML algorithms (XGBoost, Logistic Regression, Random Forest, AdaBoost, SVM) on large RNA-sequence count data.
• Utilized feature importance analysis with XGBoost to identify key genes related to TB.
• Applied pathway analysis, Gene Ontology (GO), hub-protein and protein-protein interaction (PPI) network analysis, and drug-protein interaction analysis.

Main Results:

• XGBoost achieved the highest prediction accuracy of 0.963 for TB classification.
• Identified 20 significant pathways, 24 gene ontologies, 20 hub genes, and 22 potential drugs associated with TB.
• Analysis of 100 highly expressed genes revealed complex molecular interactions.

Conclusions:

• ML models, particularly XGBoost, demonstrate high efficacy in classifying TB patients from RNA-seq data.
• The identified genes, pathways, and potential drugs offer valuable insights for TB diagnostics and therapeutic development.
• This study highlights the potential of integrating ML with multi-omics data for advancing TB research.