Machine Learning-based Diagnostic Potential of Bipolar Disorder Using Gut Microbiota Signatures

Area of Science:

• Microbiome research
• Psychiatric disorders
• Computational biology

Background:

• Bipolar disorder (BD) is a chronic illness with significant cognitive and social impairments.
• The gut microbiota-gut-brain axis is increasingly implicated in BD pathophysiology.
• Diagnostic and therapeutic strategies for BD require further development.

Purpose of the Study:

• Investigate gut microbial alterations in BD patients.
• Evaluate the diagnostic potential of gut microbiota using machine learning.
• Identify microbial biomarkers for BD noninvasive diagnosis.

Main Methods:

• 16S rRNA sequencing to analyze gut microbial composition and diversity.
• Machine learning algorithms, including random forest (RF), for classification.
• PICRUSt2 for functional prediction of microbial pathways.
• Feature selection methods to identify optimal microbial biomarkers.

Main Results:

• Reduced alpha-diversity and altered beta-diversity observed in BD patients compared to healthy controls (HC).
• RF classifier achieved high diagnostic performance (AUC=0.9316) using 35 microbial biomarkers.
• Functional analysis revealed altered pathways in neurodegeneration, lipid metabolism, and heme biosynthesis.
• Combined compositional and functional features further improved diagnostic accuracy (AUC=0.9499).

Conclusions:

• Significant compositional and functional gut microbial disturbances are present in BD.
• Machine learning-based microbiome analysis offers a promising approach for noninvasive BD diagnosis.
• Identified markers provide insights into the microbiota-gut-brain axis, supporting precision psychiatry and microbiome-targeted interventions.