Figure 1.

Scheme depicting computational steps carried out by PEMer. In PEM, when using the 454/Roche platform, randomly sheared genomic fragments are circularized and cleaved randomly into sequence stretches amenable to ultrafast sequencing (figure adapted and extended from Figure 1 in [21]). We subject resulting DNA sequences to PEMer for calling SVs relative to the reference genome ('R'). By default, PEMer uses the following processing steps: [1] construct pre-processing, [2] read-alignment, [3] optimal paired-end placement, [4] outlier-identification, [5] outlier-clustering, and [6] cluster-merging. Subsequently, [7] SVs (insertions, deletions, inversions, and more complex events) are displayed and stored in a back-end database for further analysis. In the outlier identification step, several different cutoff points C_i and C_d for the paired-end span, which are derived from the known insert-size distribution, are applied using a multi-cutoff strategy together with distinct minimally required paired-end cluster sizes N. After merging clusters constructed using different cutoff points, different PEM libraries, or different next-generation DNA sequencing platforms, an enhanced SV call resolution may be achieved.