Accuracy of SV detection across tools and variant types

Covers precision and recall performance for deletions, duplications, inversions, and insertions.

This examines the performance of various structural variation (SV) detection tools, focusing on four main SV types: deletions, duplications, inversions, and insertions. Among the evaluated tools, Manta consistently demonstrated superior accuracy in detecting deletions, achieving the highest F1 score. For insertions, both Manta and MELT showed competitive precision, while duplication and inversion detection generally exhibited lower performance across all tools. The study highlights that SV detection accuracy varies significantly depending on the type of SV and the specific tool used, underscoring the importance of choosing appropriate tools for different genomic analysis needs.

%% input data setting

sample_list='SRR622457'
input_path='D:\SV_final\_code\matlab2024_kr\data\'

addpath(genpath('D:\SV_final\_code\matlab2024_kr\data\'))

delly_file=[input_path '\delly\' sample_list '\delly.vcf']
manta_file=[input_path '\manta\' sample_list '\results\variants\diploidSV_with_INV.vcf']
gridss_file=[input_path '\gridss\' sample_list '\gridss_R.bed']
sniffle_file=[input_path '\sniffles\' sample_list '\sniffles.vcf']
wham_file=[input_path '\wham\' sample_list '\wham.vcf']
svab_file=[input_path '\svaba\' sample_list '\germline_run.svaba.sv.vcf']
lumpy_file=[input_path '\lumpy\' sample_list '\' sample_list '.vcf']
REF_input=[input_path 'TruthSet\NA12878_DGV-2016_LR-assembly_with_chr_hg38.vcf']

REF_range=make_true_table_hg38(REF_input)
REF_range_filtered=filter_function(REF_range)

delly_range=make_delly_table(delly_file)
manta_range=make_manta_table(manta_file)
sniffles_range=make_sniffles_table_2(sniffle_file)
wham_range=make_wham_table(wham_file)
gridss_range=make_gridss_table(gridss_file)
svaba_range=make_svaba_table(svab_file)
lumpy_range=make_lumpy_table(lumpy_file)

manta_range_filtered=filter_function(manta_range)
delly_range_filtered=filter_function(delly_range)
lumpy_range_filtered=filter_function(lumpy_range)
wham_range_filtered=filter_function(wham_range)
gridss_range_filtered=filter_function(gridss_range)
svaba_range_filtered=filter_function(svaba_range)
sniffles_range_filtered=filter_function(sniffles_range)


%% Tool comparision
REF_SV=REF_range_filtered

tools_vs_LSV
LSV_vs_tools

precision_test
recall_test
f1_tset

Reference for Code Development

The code implementation should refer to the following study for insights on SV detection accuracy and benchmarking:

• Joe, S., et al. Comparison of structural variant callers for massive whole-genome sequence data. BMC Genomics, 2024.