Accurate diagnosis of atopic dermatitis by combining transcriptome and microbiota data with supervised machine...

Area of Science:

• Genomics and Microbiology
• Computational Biology
• Dermatology

Background:

• Atopic dermatitis (AD) is a prevalent childhood skin condition requiring expert diagnosis.
• Emerging research links gut microbiome and host gene interactions to AD pathogenesis.
• Current diagnostic methods for AD can be complex and require specialized dermatological expertise.

Purpose of the Study:

• To create an accurate, automated diagnostic pipeline for atopic dermatitis (AD).
• To leverage host transcriptome and gut microbiota data for AD risk prediction.
• To identify novel genetic and microbial biomarkers associated with AD.

Main Methods:

• Utilized transcriptome and microbiota data from 161 subjects (AD patients and healthy controls).
• Developed and trained a machine learning classifier to predict AD risk.
• Performed feature selection to identify predictive genes and microbial taxa.

Main Results:

• The machine learning classifier achieved an average F1-score of 0.84 in differentiating AD patients from healthy individuals.
• Identified 35 predictive genes and 50 predictive microbiota features for AD.
• Discovered at least three genes and three microorganisms with direct or indirect associations with AD.

Conclusions:

• The developed omics-based pipeline demonstrates high accuracy in AD prediction.
• Identified genes and microbiota features may serve as novel biomarkers for AD.
• Further validation in independent cohorts is necessary to confirm the utility of these biomarkers.