Table 1.

The list of existing computational SV detection methods using long reads. Various methods are highlighted according to different sub-categories below. Assembly-based: de novo assembly (blue), and local assembly (orange). Alignment-based: general SV detection (grey), complex SV detection in targeted regions (yellow), and specific SV class detection (green); SV detection with combined strategy (pink). Repeat expansion: detection on long-read sequences (cyan), and detection on raw Nanopore signals (grey).

		Tool	Platform	Comments
Assembly-based SV detection	De-novo assembly	PAV35	PB, ONT	Generates phased SV callsets for haplotype-resolved assemblies from contig alignments against a reference genome.
		Dip-call141	PB	Detects large insertions and deletions from haplotype-resolved genome assemblies.
		SVIM-asm77	PB, ONT	Detects SV from diploid assemblies by pairing similar SVs from opposite haplotypes.
		SyRI79	PB, ONT	Detects SVs, as well as small variants inside rearranged regions between two genome assemblies.
		Smartie-SV87	-	Aligns contigs assembled from any type of sequencing against a reference genome.
		Assemblytics78	-	Can detect SVs, repeat expansions and contractions from contigs.
	Reference guided local assembly	PhasedSV90	PB	Creates haplotype partitioned local assemblies and supports trio assembly for accurate SV detection.
		MsPAC91	PB, ONT	Uses HMM on multiple sequence alignment of haplotype partitioned local assemblies.
		PBSV75	PB	Uses local multiple sequence realignment to detect SVs.
		SVDSS92	PB	Performs local assembly of sample-specific substrings into larger superstrings which are clustered and then used for SV detection.
Alignment-based SV detection	Rule-based	cuteSV49	PB, ONT	Uses a heuristic method to detect and genotype SVs.
		NanoSV55	PB, ONT	Uses a random forest to filter false positive SVs.
		SVIM52	PB, ONT	Uses a custom distance metric and graphs to cluster SVs and detects both tandem and interspersed duplications.
		Sniffles59	PB, ONT	Can detect complex nested SVs and estimate parameters from data set and uses NGMLR aligner.
		Sniffles260	PB, ONT	Supports somatic and population level SV calling.
		SENSV54	ONT	Uses a novel SV-aware aligner to refine breakpoints, especially for detecting long SVs (>100kbp) using low coverage ONT reads.
		PBHoney56	PB	Uses characteristics and error profile of PB sequencing.
		NanoVar51	PB, ONT	Optimized for SV detection from low-depth sequencing.
		Duet50	ONT	Incorporates SNP signatures to enable phased SV detection and genotyping.
		SKSV57	PB	Generates improved read alignment profiles for SV calling and genotyping.
		DeBreak58	PB, ONT	Identifies SVs via a density-based clustering of SV candidates obtained from alignments and uses de novo assembly detect large SVs spanning multiple reads.
		Picky53	PB, ONT	Uses a greedy seed-and-extend algorithm to improve alignment and can detect tandem duplications.
	Deep learning-based	SVision69	PB, ONT	a deep learning approach to resolve simple and complex structural variants.
		BreakNet67	PB	Predicts deletions via a CNN-LSTM deep learning model trained with feature matrices from read alignments pileup.
		MAMnet68	PB, ONT	Predicts insertions and deletions via a CNN-LSTM deep learning model trained with variant signature matrices constructed from read alignment pileups.
Ensemble Methods		NextSV71	PB	Ensemble of cuteSV2 and Sniffles2 used with minimap2 and NGMLR aligners.
		combiSV74		Combines results from six SV callers into a single call set with increased recall and precision.
Specialized SV detection	Complex SVs	SVision69	PB, ONT	Resolves complex SVs using a CNN trained on read alignment features encoded in image format.
		CORGi94	PB, ONT	Detects and visualizes complex genomic rearrangements in a local region.
		TSD95	PB	Detects and visualizes complex SVs in targeted PB deep-sequencing.
	Miscellaneous SV subtypes	rCANID96	PB, ONT	Novel element insertion detection.
		rMETL97	PB, ONT	Mobile element insertion or deletion detection.
		npInv98	PB, ONT	Non-allelic homologous recombination inversion detection.
Repeat Expansion Detection	Sequence-based	RepeatHMM101	PB, ONT	Repeat detection from long reads using HMM.
		Tandem-genotypes105	PB, ONT	Repeat detection from long reads using copy number histogram analysis.
		PacmonsTR102	PB	Repeat detection from long reads using pairHMM.
		Straglr106	PB, ONT	Scans the genome for large insertions and generates a list of coordinates and motifs which are used to genotype tandem repeats.
		adVNTR103	PB	Uses trained HMMs to genotype target variable number tandem repeats obtained with specific sequencing technologies.
		RepLong107	PB	Repeat detection from long reads using network modularity optimization.
		NanoRepeat104	PB, ONT	Repeat detection from long reads using Gaussian mixture models.
	Signal-based	STRique108	ONT	Repeat detection using Nanopore raw signals and HMM.
		NanoSatellite109	ONT	Uses squiggle-based algorithm on Nanopore raw signals.
		DeepRepeat110	ONT	Repeat detection using deep learning on Nanopore signals.
SV Genotyping		Sniffles59	PB, ONT	Computes the fraction of supporting reads for each variant against the reference and then uses allele frequency to predict genotype.
		cuteSV49	PB, ONT	Genotypes are predicted by computing the maximum likelihood of each zygosity as a function of supporting reads.
		cuteSV2142	PB, ONT	Regenotyping SVs through an accurate force-calling method.
		Samplot119	PB, ONT	Generates images with read depth and alignment information for SVs, and uses a trained ResNet-like model to predicts deletion genotypes based on these images.
		svviz2120	PB, ONT	Displays the number of supporting reads assigned to each allele, which can be used to estimate zygosity.
		SVJedi121	PB, ONT	Generates representative allele sequences for each SV and then aligns reads to these sequences to estimate allele frequencies for genotyping.
		VaPoR122	PB	Scores SV predictions by analyzing the k-mer recurrence and estimates genotype likelihood by fitting a Gaussian mixture model to the score distribution.
		LRCaller118	ONT	Alignment features are used to genotype each SV directly from long reads.
		TT-Mars123	PB	Genotypes SVs by matching their local regions to haplotype-resolved assemblies.
Somatic SV detection		Sniffles260	PB, ONT	Increases sensitivity for low-frequency SVs and additional filtering and preprocessing steps to enable non-germline SV calling.
		DeBreak58	PB, ONT	Detects non-germline SVs with clustered breakpoints in cancer genomes.
		Nanomonsv115	PB, ONT	Detects somatic SVs from paired tumor and matched control long-read sequencing data.
		SHARC116	ONT	Uses low coverage long-read sequencing to detect SVs in cancer genomes. A random forest model trained on SV features filters false positive SV calls.
		CAMPHOR117	ONT	Detects somatic SVs by comparing SVs identified from tumor samples against those from matched control samples.