Compared to the initial distribution of species in terms of cell number (first row) and biomass (second row) in the mock community. The different combinations of template (DNA or RNA/cDNA), primer pair, (Hap454 or Prym454), and read filtering strategy are indicated on the figure. Initial Filtering (IF): Initial filtering based on match with barcode and primer, and removal of reads with Ns and/or homopolymers >8 bp, and reads that are shorter than 365 bp. Quality Score (QS): requiring minimum average quality score 30 over a 50 bp moving window along the read, in addition to IF. PyroNoise (PN): the PyroNoise algorithm as implemented in mothur, with default settings. AmpliconNoise (AN): AmpliconNoise as implemented in Qiime v.1.4.0, with default settings. Chimera identification by Perseus was applied to all datasets. The filtering strategies are described in the methods section.