|
Abstract:
|
Over the past two decades, advances in sequencing technology have enabled researchers to characterize microbial diversity to an extent not previously possible. However, choice of sample-preparation and sequencing protocol is known to affect measurements arising from sequencing experiments, and no single protocol has been established as superior to others. Consequently, in microbiome studies, the relationship between measured and true microbial abundances in sampled subjects is typically unknown. To address this problem, we propose a method for deconvolving measurement error from microbiome sequencing data via observations made on specimens of known composition (i.e., “positive controls”). Specifically, we propose a semiparametric Poisson model accounting for differential detection of microbial taxa by a given protocol and also for various sources of spurious reads. We summarize performance on 16S and whole-genome sequencing datasets.
|
