Airway microbiota as a biomarker in predicting COPD and control

<bold>Background:</bold> Bacterial dysbiosis is a potential biomarker in lung disease. We hypothesized that the lower airway microbiota would classify COPD and controls more accurately than the upper airway microbiota. <bold>Methods:</bold> Oral wash (OW) and bronchoalveolar lavage (BAL) from 96 COPD patients and 97 controls from the MicroCOPD Study were analysed in Bergen, Norway. Illumina MiSeq 16SrRNA sequencing data was processed in QIIME 2 and R. <bold>Results:</bold> Lower shannon alpha-diverersity significantly predicted COPD in both OW and BAL [OR 0.57 (0.34-0.97) & 0.63 (0.41-0.96) respectively], adjusting for age, sex and smoking. The most abundant genera are shown in Figure1A. In both OW and BAL types <italic>Firmicutes</italic> was significantly more abundant in COPD (ANCOM-BC2 p-adj OW=0.02, BAL=0.04), and <italic>Alloprevotella</italic> less abundant (p-adj=0.01 in both). <italic>TM7x</italic>, <italic>Granulicatella</italic>, and <italic>Neisseria</italic> differed significantly between COPD and controls in BAL (ANCOM-BC2 p-adj<0.05). Receiver operating characteristic (ROC) analysis of random forest models of the 10 most important genera for classification showed OW outperformed BAL in classifying COPD and controls (OW ROC AUC=0.72; BAL ROC AUC=0.64) (Figure 1B). <bold>Conclusion:</bold> Our study reveals notable dysbiosis between COPD and control groups in both OW and BAL samples. OW demonstrated higher classification accuracy, though yet not suitable for clinical use. Future research should explore airway microbiota with higher resolution, as a potential biomarker. <fig><object-id>erj;64/suppl_68/PA3970/F1</object-id><object-id>F1</object-id><object-id>F1</object-id><graphic></graphic></fig>