FocalSV enables target region–based structural variant assembly and refinement using single-molecule long-read sequencing data

Table 2.

Large deletions (DELs) and insertions (INSs) (≥50 bp) calling performance across three HiFi data sets

Library Metric FocalSV (target) FocalSV (auto) PAV SVIM-asm Dipcall sawfish cuteSV SVIM pbsv Sniffles2 SKSV
DEL Hifi_L1 Precision 93.14% 93.59% 92.28% 91.70% 90.33% 93.10% 88.20% 87.33% 93.03% 83.75% 85.29%
Recall 95.36% 94.78% 93.22% 93.90% 92.61% 95.34% 88.46% 88.78% 93.44% 84.65% 85.25%
F1 94.24% 94.18% 92.75% 92.79% 91.46% 94.20% 88.33% 88.05% 93.24% 84.20% 85.27%
GT concordance 99.10% 99.05% 97.11% 97.85% 96.75% 98.37% 99.07% 98.71% 97.71% 96.56% 97.98%
Hifi_L2 Precision 92.65% 93.81% 92.04% 92.01% 90.11% 93.34% 93.83% 82.90% 92.21% 92.67% 85.81%
Recall 95.00% 95.02% 94.12% 94.56% 92.76% 95.31% 87.12% 95.40% 93.78% 94.85% 78.91%
F1 93.81% 94.41% 93.07% 93.27% 91.42% 94.31% 90.35% 88.70% 92.99% 93.74% 82.22%
GT concordance 99.00% 99.00% 97.26% 98.61% 96.86% 98.22% 97.46% 97.90% 98.19% 98.31% 97.48%
Hifi_L3 Precision 92.67% 93.71% 91.71% 91.70% 89.94% 93.15% 94.23% 75.30% 92.49% 92.81% 85.88%
Recall 95.21% 94.87% 93.49% 94.44% 92.76% 95.17% 90.43% 95.34% 93.93% 95.02% 81.44%
F1 93.92% 94.29% 92.59% 93.05% 91.33% 94.15% 92.29% 84.14% 93.20% 93.90% 83.60%
GT concordance 98.98% 99.00% 96.62% 98.59% 96.83% 98.44% 98.04% 98.34% 98.60% 98.47% 98.06%
Overall Precision 92.82% 93.70% 92.01% 91.80% 90.13% 93.20% 92.09% 81.84% 92.58% 89.74% 85.66%
Recall 95.19% 94.89% 93.61% 94.30% 92.71% 95.27% 88.67% 93.17% 93.72% 91.51% 81.87%
F1 93.99% 94.29% 92.80% 93.04% 91.40% 94.22% 90.32% 86.96% 93.14% 90.61% 83.70%
GT concordance 99.03% 99.02% 97.00% 98.35% 96.81% 98.34% 98.19% 98.32% 98.17% 97.78% 97.84%
INS Hifi_L1 Precision 90.80% 88.50% 86.30% 86.50% 72.80% 83.89% 87.50% 85.10% 90.00% 79.00% 86.10%
Recall 93.90% 93.88% 93.10% 93.10% 91.40% 94.24% 82.30% 69.60% 77.40% 76.80% 86.60%
F1 92.40% 91.11% 89.60% 89.70% 81.00% 88.76% 84.90% 76.60% 83.20% 77.90% 86.40%
GT concordance 98.10% 98.39% 95.60% 97.80% 77.80% 92.38% 99.10% 90.00% 87.70% 89.30% 98.30%
Hifi_L2 Precision 89.80% 88.32% 86.40% 86.40% 73.10% 84.75% 91.80% 61.90% 89.90% 88.30% 86.20%
Recall 93.80% 93.77% 93.30% 92.90% 91.60% 94.05% 82.40% 91.60% 76.70% 91.80% 61.70%
F1 91.70% 90.96% 89.70% 89.50% 81.30% 89.16% 86.80% 73.90% 82.80% 90.00% 71.90%
GT concordance 98.30% 98.20% 96.10% 97.70% 78.50% 92.95% 97.60% 94.00% 95.40% 97.00% 97.70%
Hifi_L3 Precision 89.60% 88.31% 86.30% 86.40% 73.00% 84.33% 91.60% 45.00% 89.00% 88.60% 87.10%
Recall 93.80% 93.88% 93.00% 93.20% 91.80% 94.04% 86.00% 92.10% 76.70% 92.80% 71.00%
F1 91.70% 91.01% 89.50% 89.60% 81.30% 88.92% 88.70% 60.50% 82.40% 90.70% 78.20%
GT concordance 98.50% 98.41% 94.90% 98.00% 78.30% 92.45% 98.40% 94.30% 96.20% 97.10% 98.60%
Overall Precision 90.07% 88.38% 86.33% 86.43% 72.97% 84.32% 90.30% 64.00% 89.63% 85.30% 86.47%
Recall 93.83% 93.84% 93.13% 93.07% 91.60% 94.11% 83.57% 84.43% 76.93% 87.13% 73.10%
F1 91.93% 91.03% 89.60% 89.60% 81.20% 88.95% 86.80% 70.33% 82.80% 86.20% 78.83%
GT concordance 98.30% 98.33% 95.53% 97.83% 78.20% 92.59% 98.37% 92.77% 93.10% 94.47% 98.20%
  • The table presents recall, precision, F1 score (in %), and genotype accuracy (measured by genotype concordance) for all benchmarked SV callers. The highest scores for recall, precision, F1, and genotype accuracy are highlighted in bold. Additionally, the overall average performance for each SV type and library is summarized at the bottom of the table. Benchmarking was conducted using Truvari with the following parameter settings: p = 0.5, P = 0.5, r = 500, and O = 0.01.

This Article

  1. Genome Res. 35: 2252-2272

Preprint Server