Through the use of replicates and modelling of flow-values, we are able to identify high frequency indel (HFI) errors that could easily be mistaken for polymorphisms in the absence of replicates. Flow-level analyses suggest that these errors do not correlate with factors implicated in over-call/under-call errors. The frequency of HFIs with respect to the reference genome will yield undesirable results for a number of applications. For example, in polymorphism detection, HFI regions will yield false positives [5]. In assembly, these regions yield unresolvable differences [2] or frame-shifts. Amplicon sequencing will suffer both from run-specific HFI errors and synchronised over-call/under-call errors; it is expected that amplicons from closely related species will be synchronised in their called flows, thus exposing them to much higher error rates as a result of noisy PICs. Given the difficulty of detecting HFIs in isolation, we recommend generating two or more datasets from the same starting DNA to resolve the majority of HFI errors (i.e. run-specific HFIs).
