2.4 Statistical analysis
Demographic data including age and sex were summarized using descriptive
methods.
Microbiome data were analyzed using QIIME2 microbiome bioinformatics
platform (31).
Sampling depth of the samples was evaluated to determine if the
within-sample diversity was fully captured: saturation of the alpha
refraction plots was inspected. Within-sample diversity was measured by
Shannon’s alpha diversity index. Influence of demographic
characteristics (age and sex) and clinical variables (phenotype of VKC,
positive skin prick test total and/or serum specific IgE, topical
therapy, type of delivery and feeding, type of diet, contact with pets,
allergy family history, living place, history of atopic dermatitis
during the first year of life, history gastrointestinal disorders) were
ascertained by Kruskal-Wallis statistical test. Pairwise comparisons
were made when more than two modalities were present, and
Benjamini-Hochberg adjustment procedure for type I error inflation due
to multiple testing was used. Between-sample diversity was measured
using Bray-Curtis index. Potential sample clusters were highlighted by
3D plots obtained from Principal Coordinates Analysis (PCoA) based on
unweighted UniFrac as a distance metric.
The analysis of composition of microbiomes (ANCOM) (32) was used to
identify differential abundant bacterial and fungal taxa between VKC and
HC at various levels of hierarchy including phylum, family, genus,
species, and OTUs. Volcano plot of the effect size difference vs. ANCOM
test statistic W was used to visualize the result of differential
abundance testing.
Results