We first asked how much of each of the three genomes was undercovered by each of the three technologies (Table 3, 1 to 9, italics), ensuring comparability by downsampling each data set to 100-fold coverage, and testing several levels of relative coverage (0.5, 0.25, 0.1 and no coverage). While modest variation was seen for E. coli on all three platforms, the results for the GC-extreme genomes were striking. For example, the fraction of the GC-poor P. falciparum genome that had relative coverage ≤0.25 (that is, four-fold undercovered or worse) ranged from 0.33% in Pacific Biosciences data (best) to 3.7% in Illumina data to 22% in Ion Torrent data (worst). In the GC-rich R. sphaeroides genome, the four-fold undercoverage fractions were 0.0071% for Pacific Biosciences (best), 0.39% for Illumina, and 36% for Ion Torrent (worst). The better performance of Pacific Biosciences is probably attributable to the lack of any amplification in their process (compare [20,21]).
