Title:Modeling complex measurement error in microbiome experiments

Abstract:The relative abundances of species in a microbiome is a scientifically important parameter to estimate given the critical role that microbiomes play in human and environmental health. However, data from artificially constructed microbiomes shows that measurement error may induce substantial bias in common estimators of this quantity. To address this, we propose a semiparametric model that accounts for common forms of measurement error in microbiome experiments. Notably, our model allows relative abundances to lie on the boundary of the simplex. We present a stable algorithm for computing parameter estimates, asymptotically valid procedures for inference in this nonstandard problem, and examples of the utility of the method. Our approach can be used to select or compare experimental protocols, design experiments with appropriate control data, analyze mixed-specimen samples, and remove across-sample contamination.