Table 1

Table 2

Table 3

Table 4

Figures at a glance

Figure 1

Figure 2

Figure 3

Table 1

Dataset

# of short reads

read length

SNV

indels

TNL

INV

Sim_Ecoli

1,472,676

100

12,822

1,035

Sim_Chr1

78,859,696

100

627,576

52,067

261

211

Sim_HG

978,158,892

100

8,003,103

666,213

2,650

2,559

SRR6062143

1,558,904,052

101

3,084,732

534,739

SRR7781445

609,362,230

151

3,084,732

534,739

SRR3440404

487,042,582

250

3,076,552

519,569

SRR6691661

638,722,972

151

3,076,552

519,569

Table 1: The benchmark datasets. Sim_Ecoli, Sim_Chr1, and Sim_HG are synthetic datasets. SRR6062143 and SRR7781445 are generated from NA12878 (HG001); SRR3440404 and SRR6691661 are generated from NA24385 (HG002). TNL: translocation; INV: inversion

Table 2

Pipeline	SNV		INDEL		Runtime (minutes)
	Precision	Recall	Precision	Recall	Runtime (minutes)
Sim_Ecoli
MapCaller	100	99.3	99.9	98.9	0.1
BWA + Freebayes	99.6	99.5	100	98.7	9.6
BWA + Mpileup	100	99.2	100	98.7	3.5
BWA + GATK	99.3	76.4	100	15.9	66.4
Sim_Chr1
MapCaller	99.9	98.5	99.8	98.0	8.3
BWA + Freebayes	99.3	99.0	99.9	98.3	611.5
BWA + Mpileup	100	98.6	99.9	97.9	277.7
BWA + GATK	99.5	78.4	99.6	17.1	3264.2
Sim_HG
MapCaller	99.2	96.9	99.1	96.3	146.1
BWA + Freebayes	99.2	92.3	99.9	91.6	6954.7
BWA + Mpileup	44.7	91.8	99.9	91.0	3414.1
BWA + GATK	99.4	71.7	99.8	14.9	36812.3

Table 2: The performance comparison on synthetic datasets. GATK producing low recall is due to high sequencing error rate (2%) and low sequencing coverage (30X). We investigate the effect of sequencing error rate and coverage in the Supplementary material

Table 3

Pipeline	SNV		INDEL		Runtime (hours)
	Precision	Recall	Precision	Recall	Runtime (hours)
SRR6062143
MapCaller	76.2	98.8	66.4	89.2	2.0
BWA + Freebayes	70.8	96.2	61.3	91.8	122.7
BWA + Mpileup	22.3	97.4	70.1	73.5	70.2
BWA + GATK	77.1	97.3	67.1	95.5	203.8
SRR7781445
MapCaller	78.6	97.3	66.2	88.3	1.5
BWA + Freebayes	72.1	96.1	58.3	95.0	122.3
BWA + Mpileup	17.8	97.4	68.0	71.4	71.8
BWA + GATK	77.3	97.2	63.9	96.0	161.8
SRR3440404
MapCaller	76.4	98.3	66.0	89.8	2.0
BWA + Freebayes	67.3	95.8	60.1	95.5	116.4
BWA + Mpileup	8.5	99.7	66.8	73.1	60.9
BWA + GATK	77.1	99.7	68.7	95.7	227.9
SRR6691661
MapCaller	77.5	97.9	60.7	91.3	1.5
BWA + Freebayes	70.4	95.8	58.1	94.8	96.5
BWA + Mpileup	16.8	99.4	66.6	76.2	49.7
BWA + GATK	78.2	99.8	63.2	99.0	157.1

Table 3: The performance comparison on real datasets.

Table 4

Data set	Method	Translocation		Inversion
		Precision	Recall	Precision	Recall
Sim_Ecoli	MapCaller	100	100	100	100
	LUMPY	100	66.7	100	100
	SVDetect	100	100	100	100
	DELLY	NA	NA	100	100
Sim_Chr1	MapCaller	100	94.3	100	87.7
	LUMPY	84.2	33.3	99.5	92.9
	SVDetect	27.2	97.7	64.8	93.8
	DELLY	NA	NA	96.9	95.7
Sim_HG	MapCaller	99.9	94.8	100	87.8
	LUMPY	87.9	35.3	100	93.7
	SVDetect	28.9	95.4	62.3	93.1
	DELLY	NA	NA	99.1	95.2

Table 4: The performance comparison on real datasets.

FIGURE 1

Figure 1: (A) Paired-end Read1 and Read2 are mapped distantly due to a translocation event in the sample genome. (B) Paired-end Read3 and Read4 are mapped with same orientation due to an inversion event in the sample genome

FIGURE 2

Figure 2: An example of position frequency matrix (PFM). Five read alignments are used to count the frequency of A, C, G and T at each position. Nucleobases in red are sequencing errors, and those in blue are SNPs

FIGURE 3

Figure 3: (A) Concordant pairs (B) discordant pairs −translocation (C) discordant pairs − inversion