For paired-end reads, steps 1-6 are repeated, however the insert size is used in conjunction with the previous read position and direction to determine the read location. Our error model also tracks how many read pairs have one read that does not align. We simulate this by generating a read that consists entirely of Ns with low quality scores, forcing it not to map. For instance, for 100 bp paired-end Illumina sequencing data, one read would contain 100 Ns, each with a quality score of 'B' or 2.
