Fig. 5

Modeled pseudogene deviation as a function of read coverage. Curves show the binomial model fits of the absolute difference in frameshifts (center) and internal stops (bottom) per Mbp relative to the representative E. coli genome for assemblies generated from simulated Illumina reads with an average Q-score of 35. Pseudogene differences generally increase rapidly with decreasing read coverage. However, the rate of change depends markedly on the combination of sequencing platform and assembler. Reported fold-coverage for 1,661,482 prokaryotic GenBank assemblies (top) imply that many publicly available assemblies were generated from reads with insufficient coverage to achieve low pseudogene error