By considering errors within their sequence-context in real data, GemSIM captured known features of Illumina and Roche/454 error profiles, thus validating our approach. GemSIM therefore facilitates independent, objective, and comparable simulation of both Roche/454 and Illumina sequencing data. Analysis of the error models created by GemSIM also provided new insights into error profiles, specifically that there were substantial differences between the error profiles of the Illumina v4 and Illumina v5 sequencing runs. While it is not clear whether this difference is due to the change in chemistry or to some other factors, it does show that sequencing runs can vary substantially, even when performed by the same sequencing provider using the same machine. Furthermore, differences between these error profiles have a substantial impact on downstream analysis, as shown by our study of SNP-calling accuracy in simulated data.
