Background Childhood asthma comprises different phenotypes with complex pathophysiology. Different asthma phenotypes evoke various clinical symptoms and vary in their responses to treatments. Methods We applied k‐means clustering algorithm of twelve objective laboratory tests among 351 asthmatic children enrolled in the Taiwanese Consortium of Childhood Asthma Study (TCCAS). We constructed gene expression profiles of peripheral blood mononuclear cells (PBMC) from children with different asthma phenotypes. Results Five distinct phenotypes of childhood asthma were identified and can be characterized by either eosinophil‐predominant or neutrophil‐predominant inflammatory characteristics. In the gene expression profile analysis, significant differences were noted for neutrophil‐predominant asthma, compared with samples from all the other asthma phenotypes. The vast majority of the differentially expressed genes in neutrophil‐predominant asthma was associated with corticosteroid response. From an independent inhaled corticosteroid (ICS) response cohort, we also found neutrophils could be activated in this severe asthma phenotype and neutrophil‐predominant asthma may be associated with corticosteroid nonresponsiveness. Conclusion Phenotype clustering of childhood asthma can be helpful to identify clinically relevant patients and reveal different inflammatory characteristics in asthmatic children. Neutrophil‐predominant asthma is the most severe asthma phenotype with poor corticosteroid response. Gene expression profile of different asthma phenotypes not only improve our knowledge of childhood asthma, but also can guide asthma precision medicine. Neutrophil‐predominant asthma is the most severe asthma phenotype with poor corticosteroid response. Five distinct phenotypes of childhood asthma identified in this study with differences in lung function, symptom frequency, healthcare utilization, percentages of eosinophils and neutrophils in peripheral blood, and serum IgE. Gene expression signature in PBMC constitutes an easier way to objectively identify corticosteroid‐resistant asthma in clinical settings.