Spatial mixed models for assessing environmental exposure effects on the microbiome

Environmental exposures such as air pollution may influence human health partly through microbiome, but statistical modeling is challenged by spatial dependence across regions and ecological correlations among taxa. We propose SpaMixed, a novel spatial mixed modeling framework for microbiome count data that incorporates both sources of dependence using conditional autoregressive priors. In simulations, SpaMixed achieved high feature selection power while maintaining low false positive rates and reduced estimation error relative to existing methods. Applied to two real studies with fine particulate matter (PM_2.5) exposures, SpaMixed identified biologically plausible PM_2.5-associated taxa, supporting its utility for environmental microbiome research.