Our tests using real data also included two types of datasets: i) high coverage single sample WGS, and ii) low coverage multiple sample WGS. First, we evaluated Pamir using WGS data at 40x coverage generated from a haploid cell line with the Illumina technology (CHM1, SRA ID: SRX652547) (Chaisson et al., 2015b). We have identified a total of 22,676 insertions that corresponds to 593.5 Kb in total, of which, 2,444 were >50bp (348 Kb total) (Table 5). Chaisson et al. (2015) also generated de novo assembly of the same genome using a long read sequencing technology (Pacific Biosciences) from the same cell line, and predicted insertions with the SMRT-SV algorithm using this dataset (Chaisson et al., 2015b). Here we used an updated call set (⁠$≥50$bp) mapped to human GRCh38 (Huddleston et al., 2016) for comparisons. Pamir showed low recall rates when compared to the long read-based SMRT-SV results (Chaisson et al., 2015b). We could identify only 488 of the 12,998 insertions detected by SMRT-SV when we consider only nearby matches (less than 10bp distance) in breakpoint predictions. One of the reasons for such discrepancy is the fact that more than half of PacBio-predicted insertions are located within various repeat regions (Table 6), and short-length Illumina reads are not sufficient to properly assemble such regions. The same effect was also observed in the original publication (Chaisson et al., 2015b), where only a handful of insertions were also identified in another assembly of the same genome that was constructed with a reference-guided methodology using both Illumina WGS and bacterial artificial chromosome datasets (Steinberg et al., 2014). We observed that approximately 45% of the insertions characterized by SMRT-SV are contain either very low (⁠$≤20%$⁠) or high (⁠$≥60%$⁠) GC%, which are known to be problematic to sequence using the Illumina platform (Benjamini and Speed, 2012; Ross et al., 2013). Additionally, we found that 14,121 out of our 22,676 predicted insertions were reported in dbSNP version 147 (Within 10 bp breakpoint resolution.).