The artificial introduction of large groups of exact or near exact duplicate reads which share the same start site was first observed in Roche 454 pyrosequencing data [8]. This ‘replicate bias’ is generally attributed to issues within the emulsion PCR (emPCR) step [9], [10]. Ion Torrent library preparation also uses emPCR, making it potentially susceptible to this bias. For each run, we evaluated the distribution of read start positions, and found no obvious deviations from the uniform distribution. Consistent with Rothberg et al. [1], we find that despite the use of emulsion PCR, PGM sequences do not exhibit replicate bias.
