To evaluate complementarity, we created mixed-technology microbial data sets for each possible platform pairing (MiSeq and Ion Torrent, Ion Torrent and Pacific Biosciences, Pacific Biosciences and MiSeq) using the previously described data sets (data sets 1 to 9). Each pairing consisted of 100-fold total coverage, composed of 50-fold randomly sampled coverage from each component technology. Then we measured the fraction of each genome that fell beneath several relative coverage thresholds, comparing those results to the undercoverage values from 100-fold 'pure' coverage from the component technologies (Table 3). If the coverage biases were complementary, we would expect that the undercoverage fractions from the mixed data sets would be smaller than those measured in the component pure data sets. This did happen in some cases. For E. coli, using a mixture of Illumina and Ion Torrent data, the two-fold undercovered fraction was 0.075%, compared to 0.54% and 0.27%, respectively, for the two technologies taken separately. Similar improvements occurred for E. coli with other platform combinations. However, for the other organisms, for the technologies tested, combining data did not reduce the overall level
