To assess the bias profile across technologies efficiently, we generated data from three microbial genomes that together span a wide range of GC base composition: Plasmodium falciparum (mean 19% GC), Escherichia coli (51%) and Rhodobacter sphaeroides (69%). All three genomes have finished reference sequences, thus facilitating a definitive analysis (see Materials and methods). Only data from Illumina (MiSeq), Ion Torrent and Pacific Biosciences were examined, not from Complete Genomics, which generates only human sequencing data. For all analyses, we note that although results are categorized by sequencing technology, in fact bias can also be introduced by library construction, and that disentanglement of these variables would require additional experiments. For the following bacterial genome analysis, Illumina libraries were made following a low-input variation of the protocol detailed in Fisher et al. [31], modified with Kapa Biosystems reagents (see Materials and methods), and both Ion Torrent and Pacific Bioscience libraries were generated using the respective manufacturers' reagents and recommended protocols (see Materials and methods).
