Table 2. Percent similarity to the respective Sanger sequence for the datasets B1 to BC7 from Maestri et al. (2019). For the mixed samples, 300 reads of each of the seven DNA barcodes were combined into a single file, from which NGSpecesID generated multiple consensus sequences. NGSpeciesID was run using Medaka polishing. * Here the Sanger sequence from Maestri et al. 2019 was shorter than the expected fragment length and all the consensus sequences. In these cases we only calculated the percentage similarity for the region covered by the respective Sanger sequence. ** The consensus sequences from Maestri et al. 2019 are missing one or two bases at the start, which could be due to a consensus calling error, or deletion of one additional base during the primer removal. For the percentage accuracy we assumed them to be incorrectly trimmed. The highest similarity scores are highlighted in bold.