Step 2: Data processing of 16S rRNA gene sequence data towards predicting metabolic functions

We processed the sequencing read dataset using the same parameters and tools as in previous section. In brief, the sequencing reads datasets were trimmed and filtered with BBDuk (v.38.12) and then denoised with DADA2 (v1.6.0) into ASVs. These ASVs were then assigned with taxonomy using QIIME2’s taxonomy classifier with Greengenes (v13.8) [39] trained on 341f-805r region. After that, MetGEMs toolbox with Core-Function was used to investigate the overall metabolic functions across all samples, hereby KO IDs and EC numbers in each sample were rank-transformed and the geometric means of KO IDs and EC numbers of each condition (i.e. atopic dermatitis and healthy samples) were computed. To detect a difference of KO IDs and EC numbers abundance between atopic dermatitis and healthy samples, the Wilcoxon rank-sum test was used.