Protocol

HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2)

Evidence concern: excluded from comparisons

Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2

2 evaluations · 40 results

Overview

An explanatory profile has not yet been reviewed for this record.

Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.

2 recorded evaluations, 40 metric rows. A comparison chart has not yet been validated for these results. The table retains the individual findings and their sources.

View coverage and remaining gaps across all benchmarks

Results

Results are available, but no reviewed comparison panel is linked in this release.

All evaluations

2 evaluations · 40 results. Different protocols are not a single leaderboard.

Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) · Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x

Sorted by Data io volume (Variant calling from BAM, whole run; HG003 ONT 10x) (lower is better). The best value in each column is highlighted. Decimals are rounded for display; each value links to the printed value and its source.

Tested configuration	Data io volume (Variant calling from BAM, whole run; HG003 ONT 10x)	Data io volume (Variant calling from BAM, whole run; HG003 ONT 20x)	Data io volume (Variant calling from BAM, whole run; HG003 ONT 30x)	Data io volume (Variant calling from BAM, whole run; HG003 ONT 40x)	Data io volume (Variant calling from BAM, whole run; HG003 ONT 50x)	Mean memory (Variant calling from BAM, whole run; HG003 ONT 10x)	Mean memory (Variant calling from BAM, whole run; HG003 ONT 20x)	Mean memory (Variant calling from BAM, whole run; HG003 ONT 30x)	Mean memory (Variant calling from BAM, whole run; HG003 ONT 40x)	Mean memory (Variant calling from BAM, whole run; HG003 ONT 50x)	Peak memory (Variant calling from BAM, whole run; HG003 ONT 10x)	Peak memory (Variant calling from BAM, whole run; HG003 ONT 20x)	Peak memory (Variant calling from BAM, whole run; HG003 ONT 30x)	Peak memory (Variant calling from BAM, whole run; HG003 ONT 40x)	Peak memory (Variant calling from BAM, whole run; HG003 ONT 50x)	Runtime (Variant calling from BAM, whole run; HG003 ONT 10x)	Runtime (Variant calling from BAM, whole run; HG003 ONT 20x)	Runtime (Variant calling from BAM, whole run; HG003 ONT 30x)	Runtime (Variant calling from BAM, whole run; HG003 ONT 40x)	Runtime (Variant calling from BAM, whole run; HG003 ONT 50x)
DeepVariant v1.10-beta GPU mode (Zheng et al. 2026)Evaluation	125 (best in column)	196	235	289	346	42,331	46,861	46,810	56,213	49,208	63,386	56,509 (best in column)	60,245	65,320	69,633	39	60	48	65	82
Clair3 v1.2.0 GPU mode (Zheng et al. 2026)Evaluation	137	175 (best in column)	222 (best in column)	263 (best in column)	307 (best in column)	10,180 (best in column)	6,803 (best in column)	10,189 (best in column)	6,260 (best in column)	8,096 (best in column)	45,935 (best in column)	56,643	57,901 (best in column)	62,871 (best in column)	63,174 (best in column)	22 (best in column)	16 (best in column)	17 (best in column)	18 (best in column)	20 (best in column)

All 40 result rows with coverage, uncertainty and sources

Filter evaluations

Tested model or methodProtocolDatasetEvaluation setupMetricEvidence origin

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configuration	Protocol and dataset	Finding	Evidence and details
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	137 Data io volume (Variant calling from BAM, whole run; HG003 ONT 10x) gigabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 10x, column 'Total data movement IO (GB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	10,180 Mean memory (Variant calling from BAM, whole run; HG003 ONT 10x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 10x, column 'Average memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	45,935 Peak memory (Variant calling from BAM, whole run; HG003 ONT 10x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 10x, column 'Peak memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	22 Runtime (Variant calling from BAM, whole run; HG003 ONT 10x) minute · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 10x, column 'Runtime (minutes)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	175 Data io volume (Variant calling from BAM, whole run; HG003 ONT 20x) gigabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 20x, column 'Total data movement IO (GB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	6,803 Mean memory (Variant calling from BAM, whole run; HG003 ONT 20x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 20x, column 'Average memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	56,643 Peak memory (Variant calling from BAM, whole run; HG003 ONT 20x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 20x, column 'Peak memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	16 Runtime (Variant calling from BAM, whole run; HG003 ONT 20x) minute · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 20x, column 'Runtime (minutes)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	222 Data io volume (Variant calling from BAM, whole run; HG003 ONT 30x) gigabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 30x, column 'Total data movement IO (GB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	10,189 Mean memory (Variant calling from BAM, whole run; HG003 ONT 30x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 30x, column 'Average memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	57,901 Peak memory (Variant calling from BAM, whole run; HG003 ONT 30x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 30x, column 'Peak memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	17 Runtime (Variant calling from BAM, whole run; HG003 ONT 30x) minute · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 30x, column 'Runtime (minutes)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	263 Data io volume (Variant calling from BAM, whole run; HG003 ONT 40x) gigabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 40x, column 'Total data movement IO (GB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	6,260 Mean memory (Variant calling from BAM, whole run; HG003 ONT 40x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 40x, column 'Average memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	62,871 Peak memory (Variant calling from BAM, whole run; HG003 ONT 40x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 40x, column 'Peak memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	18 Runtime (Variant calling from BAM, whole run; HG003 ONT 40x) minute · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 40x, column 'Runtime (minutes)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	307 Data io volume (Variant calling from BAM, whole run; HG003 ONT 50x) gigabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 50x, column 'Total data movement IO (GB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	8,096 Mean memory (Variant calling from BAM, whole run; HG003 ONT 50x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 50x, column 'Average memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	63,174 Peak memory (Variant calling from BAM, whole run; HG003 ONT 50x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 50x, column 'Peak memory (MB)'
Configuration: Clair3 v1.2.0 GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	20 Runtime (Variant calling from BAM, whole run; HG003 ONT 50x) minute · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source Clair3 v1.2.0 (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'Clair3 v1.2.0 (GPU mode)', read coverage 50x, column 'Runtime (minutes)'
Configuration: DeepVariant v1.10-beta GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	125 Data io volume (Variant calling from BAM, whole run; HG003 ONT 10x) gigabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source DeepVariant v1.10-beta (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'DeepVariant v1.10-beta (GPU mode)', read coverage 10x, column 'Total data movement IO (GB)'
Configuration: DeepVariant v1.10-beta GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	42,331 Mean memory (Variant calling from BAM, whole run; HG003 ONT 10x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source DeepVariant v1.10-beta (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'DeepVariant v1.10-beta (GPU mode)', read coverage 10x, column 'Average memory (MB)'
Configuration: DeepVariant v1.10-beta GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	63,386 Peak memory (Variant calling from BAM, whole run; HG003 ONT 10x) megabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source DeepVariant v1.10-beta (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'DeepVariant v1.10-beta (GPU mode)', read coverage 10x, column 'Peak memory (MB)'
Configuration: DeepVariant v1.10-beta GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	39 Runtime (Variant calling from BAM, whole run; HG003 ONT 10x) minute · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source DeepVariant v1.10-beta (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'DeepVariant v1.10-beta (GPU mode)', read coverage 10x, column 'Runtime (minutes)'
Configuration: DeepVariant v1.10-beta GPU mode (Zheng et al. 2026)	Protocol: HG003 ONT WGS variant-calling runtime, memory and I/O, Clair3 versus DeepVariant in GPU mode (Zheng et al. 2026 Supplementary Table 2) Dataset: GIAB HG003 whole genome, ONT R10.4.1 5 kHz (EPI2ME giab-2025.01), subsampled to 10x-50x	196 Data io volume (Variant calling from BAM, whole run; HG003 ONT 20x) gigabyte · lower Uncertainty: Not reported by the source Coverage: Not reported scored / Not reported eligible	Independent external evaluation · Source checked Evidence concern: excluded from comparisons Supplementary Table 2 runtimes differ from the article's Results for what appear to be the same HG003 comparisons. ONT at 30x and 50x: Clair3 GPU 17 and 20 minutes in the table, 16 and 15 in Results 3.1.1; DeepVariant GPU 48 and 82 in the table, 106 to 128 in the text. PacBio Revio at 30x and 50x: Clair3 GPU 12 and 12 in the table, 23 and 21 in Results 3.1.2; DeepVariant GPU 40 and 53 in the table, 93 and 118 in the text. The table's Clair3 PacBio runtimes equal the macOS M3 Ultra runtime in Results 3.1.3, not the Linux RTX 4090 runtimes. The table and the Supplementary Notes commands give DeepVariant v1.10-beta; Results 3.1 say v1.8. Supplementary Table 3 DeepVariant ONT SNP precision, recall and F1 at 10x and 20x (97.58%, 92.35%, 94.89% and 99.32%, 99.02%, 99.17%) differ from Results 3.2 (97.34%, 92.17%, 94.68% and 98.75%, 98.79%, 98.79%), while the Clair3 values agree. Results 3.2 cite Supplementary Table 2 for the HG001 to HG007 comparison, which is Supplementary Table 4. The hardware, DeepVariant version and runs behind Supplementary Tables 2 and 3 are therefore uncertain. Supplementary Table 1 agrees with Results 3.1.4 except WhatsHap ONT 30x (723 seconds in the table, 724 in the text). Table values are recorded; text values are not.Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Tables 1 to 3 (pages 2 and 3) versus article Results 3.1.1 to 3.1.4 and 3.2 Methods, coverage and source DeepVariant v1.10-beta (GPU mode), HG003 ONT runtime and memory (Zheng et al. 2026) model-execution-20261011-protocol-zheng2026-hg003-resource-ont Aggregation: Not reported Accelerated long-read variant calling with Clair3 for whole-genome sequencing; Accelerated long-read variant calling with Clair3: Supplementary Data · Supplementary Table 2 (image, page 2), sample 'HG003 WGS', platform 'ONT r10.4.1 5kHz', caller 'DeepVariant v1.10-beta (GPU mode)', read coverage 20x, column 'Total data movement IO (GB)'

Source checking is not independent reproduction. Release 2026-10-22-1b7cf027e4f6.

Methods and evaluation design

Procedure, tasks and evaluated configurations

Recorded evaluations

Each evaluation records what was tested and under which conditions.

Baseline coverage

Reference methods help show what a model adds beyond simple controls. We track a null control and a conventional method for each protocol.

0 of 2 active baseline roles have published Rewire measurements in this release. Measurements on a selected protocol do not establish coverage of an entire suite.

No execution recipe linked to this protocol. Recipe availability does not establish a completed evaluation.

Author-reported evaluations: 1
External evaluations: 1

Literature evidence is not a Rewire measurement. Executed but unpublished runs and private review status are not included.

Null control

Proposed control: requires review

Select a task-valid null control after reviewing inputs and metric

Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.

This is a suggested selection rule, not a validated method or a measured score.

Conventional reference

Proposed control: requires review

Select an upstream conventional reference after reviewing the full protocol

Protocol-specific applicability, permitted inputs, access, split, evaluator and execution requirements need review before implementation or execution.

This is a suggested selection rule, not a validated method or a measured score.

Protocol coverage CSV (gzip) · Model evaluation matrix (gzip) · Source table (gzip) · Release and checksums (gzip)

Coverage is derived from release 2026-10-22-1b7cf027e4f6. Source citations describe the original records; they do not validate an unreviewed baseline proposal. No results have been generated by this audit.

Run instructions

No runnable recipe has been reviewed for this protocol. Dataset access, model requirements, licences and compute requirements must be checked against its sources before execution.

Strengths, limitations and unresolved questions

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

0 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-10-22-1b7cf027e4f6
Property and statement	Original source and location	Review and provenance

No evidence rows match these filters. Choose another scope or clear the search.

How to read the evidence

Sources and history

Release 2026-10-22-1b7cf027e4f6 · Record review: source checked

2 source records and release history

Accelerated long-read variant calling with Clair3 for whole-genome sequencing · Original source · Bioinformatics 42(5), published 2026-04-10; PMC13141153 full-text XML
Accelerated long-read variant calling with Clair3: Supplementary Data · Original source · btag181_supplementary_data.pdf from the Europe PMC supplementary files of PMC13141153

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: model-execution-20261011-protocol-zheng2026-hg003-resource-ont

areas: dna-genomes
contexts: research
protocol: Variant calling from aligned HG003 BAM to VCF at five coverages per platform; runtime, peak and average memory and total data movement per run.
version: Supplementary Table 2
metric: runtime
limitations: One value per configuration and coverage is printed; the number of runs behind it is not stated and no spread is printed.; Author-reported: the authors develop Clair3; DeepVariant was run by them.; Hardware is not stated in Supplementary Table 2; the Methods give an NVIDIA GeForce RTX 4090 with two 12-core Intel Xeon Silver 4116 processors for the GPU benchmarks, and Results 3.1 describe the Figure 2a runs on one RTX 4090 with 32 CPU threads.; Results 3.1.1 give different runtimes for what appear to be the same comparisons (Clair3 GPU 16 minutes at 30x and 15 at 50x; DeepVariant GPU 106 to 128 minutes) and name DeepVariant v1.8, while Supplementary Table 2 prints 17 and 20 minutes for Clair3 and 48 and 82 minutes for DeepVariant v1.10-beta.; The Methods say runtime and memory were measured with the GNU time command, which reports host process memory, not GPU memory; how data movement was measured is not stated.; Accuracy is printed in a separate table (Supplementary Table 3); the tables do not say they come from the same runs.; The source has an unresolved evidence concern (Supplementary Tables 2 and 3 conflict with the Results text); see the source record.
source locator: Supplementary Table 2; Supplementary Notes; Results 3.1-3.2
memory measurement: GNU time (article Methods 'Computational performance')

Related records

Suggest a correction