Abstract:
|
The level of variability in microbiome composition, along with its composition and diversity, has been associated with conditions such as irritable bowel syndrome and preterm birth. However, existing methods to detect differences in microbiome volatility are limited, and the impact of differential volatility on downstream composition-based tests is not well understood. In particular, differential volatility is hypothesized to play the same role in longitudinal (composition-based) association tests as dispersion effects do in cross-sectional studies, reducing the power of analyses and making location effects more difficult to detect. We propose LMVolTest, a novel approach to quantifying and testing associations with microbiome volatility, and evaluate its performance in simulated and real data. We further explore the impact of differential volatility on the type 1 error and power of composition- and diversity-based tests, then assess whether adjusting for subject-level volatility mitigates impacts.
|