An assessment of the PGM sequencing technology was provided with the publication of the sequencing platform [1]. However, this analysis was conducted using quality-trimmed data (rather than raw reads) and reported accuracy based on a consensus rather than individual read accuracy. In addition, substantial changes have been made to the platform since the initial publication, including the change of flow order (Samba), quality control and base calling, which are likely to influence the error rates and biases in PGM data. Two studies comparing benchtop sequencers, including the PGM, were recently published [2], [3], both focusing on the assembly prospects of benchtop platforms on single isolate genomes. Consistent with the scope of these studies, they used the quality-trimming recommended by each distributor and did not generate sufficient replicate data to understand variability in accuracy. No study of PGM data to date has attempted to robustly evaluate error rates or explored the possible causal factors of errors at the flow-level. Here, we characterise the nature of errors produced by the PGM, including sequencing kit, chip type, machine variability, template G+C%, and chip,
