Statistical analysis
The characteristic data of all recruited HCs and asthma patients were showed as Mean ± SD, p -value < 0.05, analyzed by unpaired T test. T test and nonparametric test (Mann-Whitney U test) were used to analyze the mRNA expression and the methylation array of AREG, ATG3, E2F1, FOXO3, HDAC1, MMP2, NUF2, TGFB1 and TP53. We used the Benjamin Hochberg method to control the false discovery rate (FDR). The selection of distinguishingly expressed CpG sites was performed by Logistic regression analysis, with latent risk factors of age and gender [24]. The correlation between the percentage of methylation of candidate CpG sites and successive variables for instance FEV1%, FVC, FEV1 and PEF was assessed by Pearson’s correlation or Spearman’s correlation. ROC analysis was obtained to elucidate the accuracy of candidate DMSs or methylation change rate s in predicting asthma. For each candidate DMS, the optimal cutoff value for predicting asthma and corresponding sensitivity and specificity were defined by the maximum Youden index value (sensitivity + specificity-1) [25]. The methylation percentage of candidate DMSs or the methylation status (change or not change) were used for PCA to identify asthma. For each candidate DMS, the change in methylation status was defined by its optimal threshold [26]. The methylation change rate in each sample mainly referred to the probability that the methylation status of the candidate DMSs changed. The statistical analyses were implemented using SPSS version 22.0 (IBM Corporation, Armonk, NY, USA). A two-tailed p -value < 0.05 was considered statistically significant, **** p < 0.0001; * p < 0.05.