Comparison of the performance of Bowtie, MAQ, and segemehl on two real-life datasets.

(a) The genomic paired DNA library with 19 812 604 Illumina reads was matched using MAQ (chromosome by chromosome), Bowtie (single index), and segemehl (seven enhanced suffix arrays each representing a disjoint subset of the human chromosomes). Bowtie was used with and without quality values. To simulate a MAQ run without quality information, an average quality value was assigned to all bases of the Illumina data set. The total number of matches differs significantly: Bowtie outnumbers all other programs. segemehl still reports ten times more matches than MAQ without quality values. The number of exact matches shows a 20-fold increase. Although MAQ improves when quality values are used, the total number of matches remains small in contrast to the other programs. Differences in the number of exactly matching reads reflect the distinct handling of repetitive and uninformative reads. In segemehl, all reads matching more often than t = 500 times are dropped. (b) 38171 reads of a small short RNA library sequenced with 454 were matched to the A. thaliana genome. Compared to Bowtie and MAQ, segemehl mapped significantly more reads. Allowing for one error, segemehl matches twice as many reads as Bowtie, due to the fact that segemehl, unlike Bowtie, allows for indels. Note that segemehl discarded a few perfect matches since the corresponding seeds occur more than t = 500 times in the reference sequence.