Future improvements to GemSIM may include increasing the number of bases tracked before the current position during error model construction, as it is possible that error profiles are even more heterogenous than reported here. For Roche/454 sequencing, indel errors are known to increase with increasing homopolymer length, while there is evidence to show that Illumina sequencing accuracy can be influenced by the sequence up to 10 bases before the current position [9]. Currently, the number of bases before the current position is limited by both memory requirements and the need for the sequence-context word to be present in the control dataset. With future improvements in computing power and memory handling, it will be feasible to allow users to optionally increase the sequence-context word length, when appropriate.
