Background: Respiratory microbiome studies have improved our understanding of the various phenotypes and endotypes in heterogeneous asthma. However, the relationship between the respiratory microbiome and clinical phenotypes in children with asthma remains unclear. We aimed to identify microbiome-driven clusters reflecting the clinical features of asthma and their dominant microbiotas in children with asthma. 
Methods: Induced sputum was collected from children with asthma, and microbiome profiles were generated via sequencing of the V3-V4 region of the 16S rRNA gene. Cluster analysis was performed using the partitioning around medoid clustering method. The dominant microbiota in each cluster was determined using linear discriminant effect size analysis. Each cluster was analyzed to identify associations between the dominant microbiota, clinical phenotype, and inflammatory cytokines. 
Results: We evaluated 83 children diagnosed with asthma. Among 4 clusters reflecting the clinical characteristics of asthma, cluster 1, dominated by the genera Haemophilus and Neisseria, demonstrated lower postbronchodilator (BD) forced expiratory volume in 1 second (FEV1)/forced vital capacity (FVC) than the other clusters and more mixed granulocytic asthma. Neisseria correlated negatively with pre-BD and post-BD FEV1/FVC. Haemophilus and Neisseria correlated positively with programmed death-ligand (PD-L) 1. 
Conclusion: To our knowledge, this study is the first to analyze the relationship between an unbiased microbiome-driven cluster and clinical phenotype in children with asthma. The cluster dominated by Haemophilus and Neisseria was characterized by fixed airflow obstruction and mixed granulocytic asthma, which correlated with PD-L1 levels. Thus, unbiased microbiome-driven clustering can help identify new asthma phenotypes related to endotypes in childhood asthma.

Key words: Asthma, Children, Cluster analysis, Cytokines, Microbiota, Phenotype