ABSTRACT: Pneumonia is one of the most important causes of morbidity and mortality in children. Identification and characterization of pathogens that cause infections are crucial for accurate treatment and accelerated recovery. However, in most cases, the causative agent cannot be identified, which is partly due to the limited spectrum of pathogens covered by current diagnostics based on nucleic acid amplification. Therefore, in this study, we explored the application of metagenomic next-generation sequencing (mNGS) for the diagnosis of children with severe pneumonia. From April to July 2017, 32 hospitalized children with severe nonresponding pneumonia in Shenzhen Children's Hospital were included in this study. Blood tests were conducted immediately after hospitalization to assess cell counts and inflammatory markers, oropharyngeal swabs were collected to identify common pathogens by qPCR and culture. After bronchoscopy, bronchoalveolar lavage fluid (BALF) samples were collected for further pathogen identification using standardized diagnostic tests and mNGS. Blood tests were normal in 3 of the 32 children. In 9 oropharyngeal swabs, bacterial pathogens were detected, in 5 of these Mycoplasma pneumoniae was detected. Adenovirus was detected in 5 BALF samples, using the Direct Immunofluorescence Assay (DFA). In 15 cases, no common pathogens were found in BALF samples, using the current standard diagnostic tests, while in all 32 BALFs, pathogens were identified using mNGS, including adenovirus, Mycoplasma pneumoniae, Streptococcus pneumoniae, Haemophilus influenzae, Moraxella catarrhalis, cytomegalovirus and bocavirus. This study shows that, with mNGS, the sensitivity of detection of the causative pathogens in children with severe nonresponding pneumonia is significantly improved. In addition, mNGS gives more strain specific information, helps to identify new pathogens and could potentially help to trace and control outbreaks. In this study, we have shown that it is possible to have the results within 24 hours, making the application of mNGS feasible for clinical diagnostics.