Construction and Validation of a Machine Learning Model Based on Clinical and Microbiomic Features for Predicting...

Area of Science:

• Pulmonary Medicine
• Microbiology
• Bioinformatics

Background:

• Chronic mucus hypersecretion (CMH) is a key feature in COPD progression.
• Understanding the airway microbiome's role in CMH is crucial for disease management.

Purpose of the Study:

• To investigate clinical and airway microbiome features associated with CMH in COPD patients.
• To develop and validate a machine learning model for identifying CMH status.

Main Methods:

• 319 COPD patients were categorized into high mucus secretion (CMH) and non-high mucus secretion groups.
• Airway microbiome structure was analyzed using 16S rRNA sequencing.
• Machine learning models, including Bayesian Network (BN), were trained and tested using clinical and microbiome data.

Main Results:

• The CMH group exhibited distinct clinical characteristics and significant airway microbiome dysbiosis, with lower diversity and enrichment of pathogens like Haemophilus and Pseudomonas.
• Disease duration, Haemophilus and Pseudomonas abundance, AECOPD history, and predicted FEV1% were key predictors of CMH.
• The BN model achieved high performance (AUC=0.911) in identifying CMH status, outperforming other algorithms.

Conclusions:

• CMH in COPD is associated with airway dysbiosis and increased potentially pathogenic bacteria.
• The developed BN model demonstrates strong generalization ability for identifying CMH phenotype in COPD.