International multicenter development of ensemble machine learning driven host response based diagnosis for...

Area of Science:

• * Molecular Biology
• * Bioinformatics
• * Machine Learning in Medicine

Background:

• * Accurate diagnosis and treatment monitoring of active pulmonary tuberculosis (TB) remain significant clinical challenges.
• * Current diagnostic methods for TB can be slow and lack sensitivity, impacting patient outcomes.
• * Host gene expression signatures offer a promising avenue for developing novel diagnostic tools.

Purpose of the Study:

• * To develop and validate TB-Scope, a host-gene-expression-based machine learning model for diagnosing active pulmonary tuberculosis (ATB).
• * To create a robust diagnostic decision model integrating multiple classifiers for improved accuracy.
• * To assess the performance of TB-Scope across diverse patient populations and data platforms.

Main Methods:

• * Development of an ensemble machine learning classification model (TB-Scope) using host gene expression data.
• * Selection of 143 feature genes (biomarkers) from large-scale microarray datasets (N=1,258) for ATB prediction.
• * Optimization of the Top Scoring Pairs (TSP) ensemble classifier and construction of a multi-classifier decision model, validated on 12 independent datasets (N=1,786).

Main Results:

• * TB-Scope demonstrated high accuracy in diagnosing active pulmonary tuberculosis (ATB) across multiple validation datasets.
• * The model effectively distinguished ATB from healthy controls, latent TB infection (LTBI), and other diseases (ODs).
• * Performance was consistent across diverse data platforms (microarray and RNA-seq) and patient demographics (children and adults) from various countries.

Conclusions:

• * TB-Scope is a powerful and reliable tool for the accurate diagnosis of active pulmonary tuberculosis.
• * The host-gene-expression-based approach offers a promising strategy for improving TB diagnostics.
• * The validated model shows potential for clinical application in diverse global settings.