DeSHAPE-α: Deconvolution of quantile Structures of Heterogeneity, Asymmetry, and Pattern to advance microbial α-diversity analysis
Zihan Zhu, Yiqian Zhang, Shili Lin, Liangliang ZhangAbstract
Motivation
Alpha diversity is a core descriptor of microbial communities, yet group comparisons are typically limited to differences in mean or median values. Such analyses do not separately characterize distributional features such as between-sample dispersion and tail asymmetry. These features are descriptive summaries rather than direct measures of ecological mechanisms, and a general framework for testing them under covariate adjustment is currently lacking.
Results
We introduce DeSHAPE, a statistical framework for decomposing group differences in alpha diversity into center, dispersion and asymmetry components. In targeted, mode-matched simulation studies, DeSHAPE shows approximate type I error calibration after adjustment for the simulated covariate and substantial power gains for dispersion differences, whereas asymmetry sensitivity is lower and increases more gradually with sample size and effect size. We further apply DeSHAPE to three microbiome datasets—pancreatic cancer tumor microbiome, cardiovascular disease-associated gut microbiome, and seasonal diet-driven gut microbiome—to separate distributional components that are not identified by conventional alpha-diversity tests.
Availability and Implementation
DeSHAPE is implemented in R and freely available at https://github.com/bioscinema/DeSHAPE.