Rapidly denoising pyrosequencing amplicon reads by exploiting rank-abundance distributions

We developed a fast method for denoising pyrosequencing for community 16S rRNA analysis. We observe a 2–4 fold reduction in the number of observed OTUs (operational taxonomic units) comparing denoised with non-denoised data. ~50,000 sequences can be denoised on a laptop within an hour, two orders of magnitude faster than published techniques. We demonstrate the effects of denoising on alpha and beta diversity of large 16S rRNA datasets.