2.2. WGCNA

The weighted gene coexpression network analysis was performed using WGCNA R package (v1.69) in the COPD and normal samples [25]. A total of 35 macrophage polarization genes (Supplementary Table 1) from the MsigDB database (https://www.gsea-msigdb.org/gsea/msigdb) were considered different traits to investigate the coexpressed genes related to macrophage polarization genes. Soft thresholding was then applied by raising correlation values to a power of 14 to amplify disparity between strong and weak correlations. The soft thresholding power was chosen to achieve approximately scale-free network topology, as recommended for biological networks [26, 27]. The resulting signed adjacency matrix was used to calculate topological overlap matrix (TO), which was then hierarchically clustered with (1-TO) as a distance measure.

Genes were then assigned into coexpression modules by dynamic tree cutting algorithm requiring minimal module size of 100 genes [28]. The modules with highly correlated eigengenes (correlation above 0.6) were merged. Module eigengene (ME) is the first principal component of the gene expression values within a module and is used to summarize the module's expression. The Pearson correlation between each gene and ME was then calculated. This value is called module membership (MM) and represents how close a particular gene is to a module. Finally, each gene was assigned to a module for which it had the highest MM. The module with the highest absolute value of the correlation coefficient with the traits was chosen as the key module for subsequent analysis.