Skin microbiome composite features in Atopic Dermatitis via integration analysis across cohorts

Abstract

Disruption of the skin microbiota is closely associated with the onset and progression of Atopic Dermatitis (AD). However, inconsistencies across studies have hindered a comprehensive understanding of their role in AD and their potential as reliable diagnostic biomarkers. To address this, we conducted a cross-cohort integrative analysis (CCIA) of raw 16S rRNA sequencing data and metadata from the largest available dataset to date, encompassing 1,522 samples across 10 independent studies. We identified consistent microbial signatures distinguishing AD patients from healthy controls. Significant alterations were observed in both -diversity and community composition between AD and control groups, while lesional and non-lesional sites within AD patients showed no significant differences. Given the impact of confounding factors on the skin microbiota, we applied MMUPHin framework to correct for batch effects and performed subgroup analyses based on different batches. Differential taxa were identified using Permutation testing, Wilcoxon rank-sum tests, and LEfSe analysis. These features were used to develop predictive models with four machine learning algorithms, achieving high diagnostic accuracy in an independent validation cohort (AUROC = 0.83). Our study provides a comprehensive reference of skin microbial alterations in AD, offering valuable insights into host-microbe interactions and highlighting their potential as diagnostic biomarkers for early detection and targeted therapeutic strategies.