2.5 | Statistical analysis
Student’s t-test, non-parametric Mann-Whitney U test, Chi-square test, and Fisher’s exact test were performed to investigate maternal clinical characteristics using R programming. The Kruskal-Wallis test and Post hoc comparisons were applied to compare neonatal clinical characteristics between DCDA-C, DCDA-D-L, and DCDA-D-S. The Uniform Manifold Approximation and Projection (UMAP) was used to compare the hair metabolome profiles between three twin groups using uMAP and ggplot2 R-packages. The hair metabolite profiles were adjusted by log transformation and Pareto scaling to ensure the Gaussian distribution of the hair dataset prior to statistical analysis. The adjusted logistic regression models were applied to eliminate the confounding effect of gestation age. Metabolic pathway activities were estimated using identified metabolites by Metaboanalyst 5.0. Moreover, the metabolic network was in silico reconstruction based on KEGG metabolic framework using the Metascape package in Cytoscape (Version 3.9.1). Pearson’s correlation was executed to determine the correlation between hair metabolite levels, metabolic pathways, physical development indices, and five developmental domains of ASQ-3. The metabolic heatmaps, pathway activity plots, and circus plots were illustrated by pheatmap, ggplot2, and GOplot R-packages.