Abstract
Bar-coded pyrosequencing has been increasingly used due to its fine taxonomic resolution and high throughput. Yet, concerns arise regarding the reproducibility of bar-coded pyrosequencing. We evaluated the run-to-run variation of bar-coded pyrosequencing in detecting bacterial community shifts and taxa dynamics. Our results demonstrate that pyrosequencing is reproducible in evaluating community shifts within a run, but not between runs. Also, the reproducibility of pyrosequencing in detecting individual taxa increased as a function of taxa abundance. Based on our findings: (1) for studies with modest sequencing depth, it is doubtful that data from different pyrosequencing runs can be considered comparable; (2) if multiple pyrosequencing runs are needed to increase the sequencing depth, additional sequencing efforts should be applied to all samples, rather than to selected samples; (3) if pyrosequencing is used for estimating bacterial population dynamics, only the abundant taxa should be considered; (4) for less-abundant taxa, the sequencing depth should be increased to ensure an accurate evaluation of taxon variation trends across samples.