Table 2.

Summary of SV genotyping methods

Category	cuteSV	LRcaller	Sniffles	SVJedi	VaPoR
Version	1.0.11	0.2	1.0.12a	1.1.0	NA
Aligner	minimap2, NGMLR	minimap2, NGMLR	minimap2, NGMLR	minimap2	minimap2, NGMLR
SV type	INSs, DELs, DUPs, INVs, TRAs	INSs, DELs, DUPs, INVs, TRAs	INSs, DELs, DUPs, INVs, TRAs	INSs, DELs, INVs, TRAs	INSs, DELs, DUPs, INVs
Data type	CLR, CCS, ONT	CLR, CCS, ONT	CLR, CCS, ONT	CLR, ONT	CLR, CCS, ONT
Input	REF, BAM, VCF	REF, BAM, VCF	BAM, VCF	REF, VCF, FASTA/FASTQ; PAF, VCF	REF, BAM, VCF/BED
Output	1 genotype	5 genotypes	1 genotype	1 genotype	1 genotype

REF: the human reference genome; BAM: alignment in BAM (binary alignment map) format; VCF: targeted SVs in VCF (variant call format); FASTA/FASTQ: sequencing data in FASTA or FASTQ format; PAF: alignment in PAF (pairwise mapping format); BED: targeted SVs in BED (browser extensible data) format. “NA” indicates the data are not available. LRcaller employs five different genotyping models: direct (AD), variant alignment (VA), joint (J), presence (PR), and reference aware variant alignment (VAr), resulting in five genotypes [32]. We used the genotypes of the default joint model when comparing it with other SV genotyping methods