All PIC terms were found to significantly contribute to the mean, with some positions shifting the mean by −0.10 to +0.05, however the fitted effects were not consistent across PIC corresponding to the same nucleotide ( Figure 6a , Table S1). This may explain why obvious groupings of nucleotides (GC versus AT, pyrmidines versus purines) as a factor in the DGLM were not significant. However, the three largest effects for a PIC are attributed to T or A flows, and the magnitude of these effects result in indel error rates up to double that of other PIC ( Figure 6b ). This suggests that sequencing of low G+C% species will result in a higher error rate than high G+C% species, consistent with the observation that S. tokodaii had a higher error rate than B. amyloliquefaciens ( Figure 3 ). While PIC 10 and 12 had substantial mean-effects, the contribution of these factors to the dispersion was less than fitted for other PICS that had substantially smaller mean-effects. Thus, these PICs consistently introduce a shift to the mean flow-value by −0.10 and +0.05 for PICS 10 and 12 respectively.
