Chronic obstructive pulmonary disease (COPD) is a lung disease characterized by airway limitation and persistent respiratory symptoms. Tobacco smoking and exposure to air pollution have been reported as major risk factors of COPD, leading to diverse clinical presentations of COPD with unclear pathogenesis and no efficient treatments are available for now. Identification of biomarkers for stratification of COPD phenotypes will benefit the discovery of potential treatment strategies. In this study, we develop a data-independent acquisition mass spectrometry (DIA-MS) platform assisted by an in-house serum peptide spectral library for in-depth profiling of serum proteome in COPD. Individualized proteomic analyses of serum samples from 88 COPD patients identified 1,062 proteins, of which 37 down-regulated and 81 up-regulated proteins involved in nuclear receptor signaling, cytokine signaling and innate immune system pathways were filtered in the late-stage COPD. A number of stage-dependent proteins was discovered; the expression levels of ADIPOQ and FAM20C are negatively correlated with forced expiratory volume (FEV1%), while CNTN3, MEGF8, CFHR2 and DAG1 are positively correlated (p<0.05). Furthermore, the differentially expressed serum proteins in smoker COPD are associated with regulation of cell migration and epithelial cell differentiation. The clustering analysis of serum proteome profiles identified two molecular subtypes in the 88 COPD patients, however, no significant differences were found in clinical characteristics between the two clusters of patients. These data suggest a diverse pathogenesis mechanism in different phenotypes of COPD patients. Based on this finding, we will further extract a reliable serum protein panel that could offer efficient stratification of COPD for clinical utility and explore potential treatment strategies toward each phenotype of COPD.