[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EE000BYPP2M8E2Y7SB32RQ":3,"skill-same:01M7EE000BYPP2M8E2Y7SB32RQ":62},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":30,"commitSha":31,"skillDirectory":32,"forksCount":14,"repositoryHtmlUrl":33,"repositoryHomepageUrl":33,"ownerHtmlUrl":34,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":35,"install":36,"resources":49},"01M7EE000BYPP2M8E2Y7SB32RQ","bio-long-read-sequencing-clair3-variants",null,"Calls germline small variants (SNPs and indels) from Oxford Nanopore and PacBio HiFi long reads with Clair3, a two-stage (pileup + full-alignment) deep-learning caller, selecting the chemistry- and basecaller-version-matched model, enabling read-based phasing, and benchmarking against GIAB with stratification. Covers why the model string is the experiment (no auto-detection, silent degradation on mismatch), why ONT homopolymer\u002FSTR indels are the residual error whole-genome F1 hides, and the somatic\u002Ftrio\u002FRNA boundary to the ClairS\u002FClair3-Trio family. Use when calling germline SNVs\u002Findels from ONT or HiFi BAMs, choosing a Clair3 model, phasing variants, or benchmarking long-read calls.","Mr-Nobody-Anonymous\u002FAll-skills","Mr-Nobody-Anonymous","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F184253119?v=4","01M7AEXFJHKTYQW5R0ZCQQTWFV","anonymous","01M7C00M3GQR4QKP9GZA42GFAK",0,"MIT","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-long-read-sequencing-clair3-variants\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: bio-long-read-sequencing-clair3-variants\ndescription: Calls germline small variants (SNPs and indels) from Oxford Nanopore and PacBio HiFi long reads with Clair3, a two-stage (pileup + full-alignment) deep-learning caller, selecting the chemistry- and basecaller-version-matched model, enabling read-based phasing, and benchmarking against GIAB with stratification. Covers why the model string is the experiment (no auto-detection, silent degradation on mismatch), why ONT homopolymer\u002FSTR indels are the residual error whole-genome F1 hides, and the somatic\u002Ftrio\u002FRNA boundary to the ClairS\u002FClair3-Trio family. Use when calling germline SNVs\u002Findels from ONT or HiFi BAMs, choosing a Clair3 model, phasing variants, or benchmarking long-read calls.\ntool_type: cli\nprimary_tool: Clair3\nsource: \"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills\"\nsource_repository: \"GPTomics\u002FbioSkills\"\nsource_path: \"long-read-sequencing\u002Fclair3-variants\u002FSKILL.md\"\nlicense: \"MIT\"\nimported_at: \"2026-09-20\"\n---\n\n## Version Compatibility\n\nReference examples tested with: Clair3 2.0+, whatshap 2.0+, bcftools 1.19+, hap.py 0.3.15+.\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- CLI: `\u003Ctool> --version` then `\u003Ctool> --help` to confirm flags\n\nResults depend on inputs that outlive the binary version - record them:\n- The Clair3 MODEL must match the platform + chemistry + basecaller tier + basecaller version (e.g. `r1041_e82_400bps_sup_v500`). There is NO auto-detection; `--model_path` is mandatory and a mismatch silently degrades calls.\n- Clair3 v2 moved TensorFlow -> PyTorch; models are `pileup.pt`\u002F`full_alignment.pt`. v1 TensorFlow models do NOT load in v2.\n- The full ONT model set (every version, hac\u002Ffast, `_with_mv` signal-aware) lives in the rerio `clair3_models\u002F` repo; only a subset is bundled.\n\nIf code throws an error, introspect the installed tool (`run_clair3.sh --help`) and adapt the example to the actual API rather than retrying.\n\n# Clair3 Variant Calling\n\n**\"Call variants from my long reads\"** -> Run Clair3 with the model that matches how the reads were basecalled, phase, and benchmark with stratification - because the model string, not the command, determines accuracy.\n- CLI: `run_clair3.sh --bam_fn=aln.bam --ref_fn=ref.fa --output=out\u002F --threads=16 --platform=ont --model_path=\u002Fmodels\u002Fr1041_e82_400bps_sup_v500`\n\nScope: germline diploid SNPs + small indels. NOT structural variants (-> structural-variants), NOT somatic\u002Fmosaic (-> ClairS\u002FClairS-TO), NOT RNA (-> Clair3-RNA).\n\n## The Single Most Important Modern Insight -- The Model String Is the Experiment, and ONT Indels Hide in the Strata\n\nClair3's accuracy is gated by two facts a naive user misses:\n\n1. **The model is hand-picked and a mismatch fails silently.** There is no auto-detection - the user must point `--model_path` at a specific model folder. Three axes must ALL match: chemistry (`r941` vs `r1041`), basecaller tier (`fast`\u002F`hac`\u002F`sup`), and basecaller version (`g5014`\u002F`v430`\u002F`v500`\u002F`v520`), plus the optional `_with_mv` signal-aware axis if the BAM has Dorado `mv` tags. Wrong model = no crash, no warning, measurably worse calls (indels most). Derive the model from the basecaller string in the run metadata; pick the model version closest to but not above the basecaller version.\n2. **ONT indels in homopolymers\u002FSTRs are the residual error whole-genome F1 conceals.** Even on R10.4.1 sup, insertions\u002Fdeletions in homopolymer runs and short tandem repeats are the weak point (G\u002FC homopolymers worst), because the pore cannot reliably count identical consecutive bases. A genome-wide indel F1 of ~99.5% hides much lower performance inside LowComplexity\u002Fhomopolymer strata - exactly the medically relevant loci. HiFi largely solves this; do not transfer ONT-indel pessimism to HiFi. Always benchmark with GIAB stratification, never a single global number.\n\n## Two-Stage Architecture\n\nClair3 \"symphonizes\" two networks: a fast **pileup model** (summarized per-position statistics) that calls the large majority of sites, and a slow **full-alignment model** (haplotype-resolved read tensor) that re-evaluates only the uncertain subset. Internally Clair3 phases the top het-SNP pileup calls with WhatsHap, haplotags the BAM, and feeds the haplotagged reads to the full-alignment model - which is why read-based phasing buys ~6% indel F1, not cosmetics. Output: `merge_output.vcf.gz` (final).\n\n## Model Selection\n\nModel name anatomy (`r1041_e82_400bps_sup_v500`): pore (`r1041`=R10.4.1), flowcell (`e82`), speed (`400bps`), basecaller tier (`sup`\u002F`hac`\u002F`fast`), basecaller version (`v500`=Dorado 5.0.0, `g5014`=Guppy 5.0.14). The `_with_mv` suffix uses Dorado move-table tags for best accuracy when present.\n\n| Data | `--platform` | Model |\n|------|--------------|-------|\n| ONT R10.4.1 sup, Dorado v5.x, mv tags present | ont | `r1041_e82_400bps_sup_v520_with_mv` |\n| ONT R10.4.1 sup, Dorado v5.0.0 | ont | `r1041_e82_400bps_sup_v500` |\n| ONT R10.4.1 hac | ont | `r1041_e82_400bps_hac_v500`\u002F`_v520` |\n| ONT R9.4.1 (any tier) | ont | `r941_prom_sup_g5014` |\n| PacBio HiFi Revio | hifi | `hifi_revio` |\n| PacBio HiFi Sequel II | hifi | `hifi_sequel2` |\n| Illumina (supported) | ilmn | `ilmn` |\n| PacBio CLR | - | not supported -> PEPPER-Margin-DeepVariant |\n\n## Decision Tree by Scenario\n\n| Scenario | Tool | Why |\n|----------|------|-----|\n| Germline SNV\u002Findel, single sample | Clair3 | this skill |\n| Somatic, paired tumor-normal | ClairS | VAF-aware; Clair3 germline priors cannot find low-VAF somatic |\n| Somatic, tumor-only | ClairS-TO | tumor-only ensemble |\n| De novo \u002F Mendelian trio | Clair3-Nova \u002F Clair3-Trio | family-aware |\n| Long-read RNA variants | Clair3-RNA | RNA model |\n| ONT R10.4.1, also considering DeepVariant | either | neck-and-neck on R10 sup; native-ONT DeepVariant (Kolesnikov 2024) superseded PEPPER-Margin |\n| Non-human \u002F draft \u002F bacterial reference | Clair3 + `--include_all_ctgs` | default calls only chr1-22,X,Y -> empty output otherwise |\n| Cohort joint genotyping | Clair3 gVCF -> GLnexus | `bcftools merge` on gVCFs is NOT joint genotyping |\n\n## Core Commands\n\n```bash\n# Germline ONT calling (model MUST match the basecaller)\nrun_clair3.sh \\\n  --bam_fn=aln.bam --ref_fn=ref.fa --output=clair3_out\u002F \\\n  --threads=16 --platform=ont \\\n  --model_path=\u002Fopt\u002Fmodels\u002Fr1041_e82_400bps_sup_v500\n# final VCF: clair3_out\u002Fmerge_output.vcf.gz\n\n# Phase the final output VCF (WhatsHap); --longphase_for_phasing swaps only the INTERNAL\n# phaser to LongPhase (faster, SV-aware). For a LongPhase-phased final VCF use\n# --use_longphase_for_final_output_phasing instead of --enable_phasing.\nrun_clair3.sh ... --enable_phasing --longphase_for_phasing\n# Phased calls go to clair3_out\u002Fphased_merge_output.vcf.gz; merge_output.vcf.gz stays UNPHASED.\n\n# Non-human \u002F draft assembly reference - call ALL contigs\nrun_clair3.sh ... --include_all_ctgs\n\n# Targeted \u002F amplicon panel\nrun_clair3.sh ... --bed_fn=panel.bed --gvcf\n\n# Benchmark against GIAB with stratification (the step that reveals ONT indel errors)\nhap.py giab_truth.vcf.gz clair3_out\u002Fmerge_output.vcf.gz \\\n  -f giab_confident.bed -r ref.fa --engine=vcfeval \\\n  --stratification giab_stratifications.tsv -o bench\u002Fhg002\n```\n\n## Per-Method Failure Modes\n\n### Silent model mismatch\n**Trigger:** `--model_path` pointing at a model that does not match the basecaller chemistry\u002Ftier\u002Fversion. **Mechanism:** no auto-detection; the wrong network runs. **Symptom:** no error, lower F1 (indels most). **Fix:** derive the model from the basecaller string; verify the folder exists (rerio for the full set); for v2 ensure `.pt` models.\n\n### Clair3 found nothing on a non-human reference\n**Trigger:** bacterial genome or draft assembly without chr1-22,X,Y names. **Mechanism:** Clair3 calls only standard human contigs by default. **Symptom:** near-empty VCF. **Fix:** `--include_all_ctgs`.\n\n### Global F1 looks great, clinical genes are wrong\n**Trigger:** reporting only whole-genome F1. **Mechanism:** ONT indel errors concentrate in homopolymer\u002FSTR\u002Flow-complexity strata. **Symptom:** ~99.5% global indel F1 but much lower in LowComplexity. **Fix:** stratify with GIAB BEDs (Dwarshuis 2024); use CMRG for medically relevant genes.\n\n### Treating Clair3 as a somatic caller\n**Trigger:** lowering `--snp_min_af`\u002F`--indel_min_af` to catch low-VAF variants. **Mechanism:** germline model expects ~0.5\u002F1.0 allele fractions, is not VAF-aware. **Symptom:** germline-model false positives at low AF, missed true somatic. **Fix:** ClairS (paired) \u002F ClairS-TO (tumor-only).\n\n### v1 model with v2 Clair3\n**Trigger:** an old TensorFlow model dir with Clair3 v2. **Mechanism:** v2 needs PyTorch `.pt` models. **Symptom:** model load failure. **Fix:** use `pileup.pt`\u002F`full_alignment.pt` models (Converted Rerio).\n\n## Quantitative Thresholds\n\n| Threshold | Source | Rationale |\n|-----------|--------|-----------|\n| Recommended depth ~20-60x | Clair3 guidance | sensitivity (hets, indels) falls off below ~20x; `--min_coverage` default 2 is a floor, not a recommendation |\n| Phasing buys ~6% indel F1 | Zheng 2022 | haplotagged reads disambiguate indel alleles in repeats |\n| ONT R10.4.1 sup: SNP F1 ~99.99%, indel F1 ~99.5% | GIAB benchmarks | indel residual lives in homopolymer\u002FSTR strata |\n| `--var_pct_full` 0.3 (default) | Clair3 README | fraction of low-quality pileup calls re-run by full-alignment; raise for recall, slower |\n| Stratify with GIAB \u002F CMRG | Dwarshuis 2024 | global F1 hides the ONT indel problem |\n\n## Common Errors\n\n| Error \u002F symptom | Cause | Solution |\n|-----------------|-------|----------|\n| Empty\u002Fnear-empty VCF on non-human ref | default calls only chr1-22,X,Y | `--include_all_ctgs` |\n| Model fails to load | v1 TF model with v2 Clair3 | use `.pt` (PyTorch) models |\n| `--model_path ...\u002Fmodels\u002Font` not found | no generic `ont`\u002F`hifi` model | point at a specific model subfolder |\n| Worse-than-expected indels | wrong-version or wrong-tier model | match the basecaller model exactly |\n| \"joint genotyping\" gave odd merges | `bcftools merge` on gVCFs is not joint calling | use GLnexus |\n| Looking for somatic\u002Flow-VAF variants | germline caller | use ClairS \u002F ClairS-TO |\n\n## References\n\n- Zheng Z, Li S, Su J, Leung AWS, Lam TW, Luo R. 2022. Symphonizing pileup and full-alignment for deep learning-based long-read variant calling (Clair3). *Nat Comput Sci* 2:797-803.\n- Zheng Z, He M, Yu X, et al. 2026. Accelerated long-read variant calling with Clair3 for whole-genome sequencing. *Bioinformatics* (advance access) btag181.\n- Kolesnikov A, Cook D, Nattestad M, et al. 2024. Local read haplotagging enables accurate long-read small variant calling. *Nat Commun* 15:5907.\n- Dwarshuis N, Kalra D, McDaniel J, et al. 2024. The GIAB genomic stratifications resource for human reference genomes. *Nat Commun* 15:9029.\n- Lin JH, Chen LC, Yu SC, Huang YT. 2022. LongPhase: an ultra-fast chromosome-scale phasing algorithm for small and large variants. *Bioinformatics* 38(7):1816-1822.\n- Chen L, Zheng Z, Su J, et al. 2025. ClairS-TO: a deep-learning method for long-read tumor-only somatic small variant calling. *Nat Commun* 16:9630.\n\n## Related Skills\n\n- basecalling - The basecaller model+version the Clair3 model must match\n- long-read-alignment - Produces the BAM (keep `--MD`; use minimap2 >=2.28)\n- haplotype-phasing - whatshap\u002Flongphase phasing and haplotagging Clair3 uses internally\n- medaka-polishing - ONT consensus; medaka diploid variant calling is deprecated in favor of Clair3\n- structural-variants - SVs are out of Clair3's scope (Sniffles2\u002FcuteSV)\n- variant-calling\u002Fdeepvariant - DeepVariant native ONT\u002FHiFi models (neck-and-neck on R10)\n- variant-calling\u002Fvcf-statistics - Summarize\u002Ffilter the VCF Clair3 emits\n- clinical-databases\u002Fvariant-prioritization - Prioritize the called variants\n","## Version Compatibility\n\nReference examples tested with: Clair3 2.0+, whatshap 2.0+, bcftools 1.19+, hap.py 0.3.15+.\n\nBefore using code patterns, verify installed versions match. If versions differ:\n- CLI: `\u003Ctool> --version` then `\u003Ctool> --help` to confirm flags\n\nResults depend on inputs that outlive the binary version - record them:\n- The Clair3 MODEL must match the platform + chemistry + basecaller tier + basecaller version (e.g. `r1041_e82_400bps_sup_v500`). There is NO auto-detection; `--model_path` is mandatory and a mismatch silently degrades calls.\n- Clair3 v2 moved TensorFlow -> PyTorch; models are `pileup.pt`\u002F`full_alignment.pt`. v1 TensorFlow models do NOT load in v2.\n- The full ONT model set (every version, hac\u002Ffast, `_with_mv` signal-aware) lives in the rerio `clair3_models\u002F` repo; only a subset is bundled.\n\nIf code throws an error, introspect the installed tool (`run_clair3.sh --help`) and adapt the example to the actual API rather than retrying.\n\n# Clair3 Variant Calling\n\n**\"Call variants from my long reads\"** -> Run Clair3 with the model that matches how the reads were basecalled, phase, and benchmark with stratification - because the model string, not the command, determines accuracy.\n- CLI: `run_clair3.sh --bam_fn=aln.bam --ref_fn=ref.fa --output=out\u002F --threads=16 --platform=ont --model_path=\u002Fmodels\u002Fr1041_e82_400bps_sup_v500`\n\nScope: germline diploid SNPs + small indels. NOT structural variants (-> structural-variants), NOT somatic\u002Fmosaic (-> ClairS\u002FClairS-TO), NOT RNA (-> Clair3-RNA).\n\n## The Single Most Important Modern Insight -- The Model String Is the Experiment, and ONT Indels Hide in the Strata\n\nClair3's accuracy is gated by two facts a naive user misses:\n\n1. **The model is hand-picked and a mismatch fails silently.** There is no auto-detection - the user must point `--model_path` at a specific model folder. Three axes must ALL match: chemistry (`r941` vs `r1041`), basecaller tier (`fast`\u002F`hac`\u002F`sup`), and basecaller version (`g5014`\u002F`v430`\u002F`v500`\u002F`v520`), plus the optional `_with_mv` signal-aware axis if the BAM has Dorado `mv` tags. Wrong model = no crash, no warning, measurably worse calls (indels most). Derive the model from the basecaller string in the run metadata; pick the model version closest to but not above the basecaller version.\n2. **ONT indels in homopolymers\u002FSTRs are the residual error whole-genome F1 conceals.** Even on R10.4.1 sup, insertions\u002Fdeletions in homopolymer runs and short tandem repeats are the weak point (G\u002FC homopolymers worst), because the pore cannot reliably count identical consecutive bases. A genome-wide indel F1 of ~99.5% hides much lower performance inside LowComplexity\u002Fhomopolymer strata - exactly the medically relevant loci. HiFi largely solves this; do not transfer ONT-indel pessimism to HiFi. Always benchmark with GIAB stratification, never a single global number.\n\n## Two-Stage Architecture\n\nClair3 \"symphonizes\" two networks: a fast **pileup model** (summarized per-position statistics) that calls the large majority of sites, and a slow **full-alignment model** (haplotype-resolved read tensor) that re-evaluates only the uncertain subset. Internally Clair3 phases the top het-SNP pileup calls with WhatsHap, haplotags the BAM, and feeds the haplotagged reads to the full-alignment model - which is why read-based phasing buys ~6% indel F1, not cosmetics. Output: `merge_output.vcf.gz` (final).\n\n## Model Selection\n\nModel name anatomy (`r1041_e82_400bps_sup_v500`): pore (`r1041`=R10.4.1), flowcell (`e82`), speed (`400bps`), basecaller tier (`sup`\u002F`hac`\u002F`fast`), basecaller version (`v500`=Dorado 5.0.0, `g5014`=Guppy 5.0.14). The `_with_mv` suffix uses Dorado move-table tags for best accuracy when present.\n\n| Data | `--platform` | Model |\n|------|--------------|-------|\n| ONT R10.4.1 sup, Dorado v5.x, mv tags present | ont | `r1041_e82_400bps_sup_v520_with_mv` |\n| ONT R10.4.1 sup, Dorado v5.0.0 | ont | `r1041_e82_400bps_sup_v500` |\n| ONT R10.4.1 hac | ont | `r1041_e82_400bps_hac_v500`\u002F`_v520` |\n| ONT R9.4.1 (any tier) | ont | `r941_prom_sup_g5014` |\n| PacBio HiFi Revio | hifi | `hifi_revio` |\n| PacBio HiFi Sequel II | hifi | `hifi_sequel2` |\n| Illumina (supported) | ilmn | `ilmn` |\n| PacBio CLR | - | not supported -> PEPPER-Margin-DeepVariant |\n\n## Decision Tree by Scenario\n\n| Scenario | Tool | Why |\n|----------|------|-----|\n| Germline SNV\u002Findel, single sample | Clair3 | this skill |\n| Somatic, paired tumor-normal | ClairS | VAF-aware; Clair3 germline priors cannot find low-VAF somatic |\n| Somatic, tumor-only | ClairS-TO | tumor-only ensemble |\n| De novo \u002F Mendelian trio | Clair3-Nova \u002F Clair3-Trio | family-aware |\n| Long-read RNA variants | Clair3-RNA | RNA model |\n| ONT R10.4.1, also considering DeepVariant | either | neck-and-neck on R10 sup; native-ONT DeepVariant (Kolesnikov 2024) superseded PEPPER-Margin |\n| Non-human \u002F draft \u002F bacterial reference | Clair3 + `--include_all_ctgs` | default calls only chr1-22,X,Y -> empty output otherwise |\n| Cohort joint genotyping | Clair3 gVCF -> GLnexus | `bcftools merge` on gVCFs is NOT joint genotyping |\n\n## Core Commands\n\n```bash\n# Germline ONT calling (model MUST match the basecaller)\nrun_clair3.sh \\\n  --bam_fn=aln.bam --ref_fn=ref.fa --output=clair3_out\u002F \\\n  --threads=16 --platform=ont \\\n  --model_path=\u002Fopt\u002Fmodels\u002Fr1041_e82_400bps_sup_v500\n# final VCF: clair3_out\u002Fmerge_output.vcf.gz\n\n# Phase the final output VCF (WhatsHap); --longphase_for_phasing swaps only the INTERNAL\n# phaser to LongPhase (faster, SV-aware). For a LongPhase-phased final VCF use\n# --use_longphase_for_final_output_phasing instead of --enable_phasing.\nrun_clair3.sh ... --enable_phasing --longphase_for_phasing\n# Phased calls go to clair3_out\u002Fphased_merge_output.vcf.gz; merge_output.vcf.gz stays UNPHASED.\n\n# Non-human \u002F draft assembly reference - call ALL contigs\nrun_clair3.sh ... --include_all_ctgs\n\n# Targeted \u002F amplicon panel\nrun_clair3.sh ... --bed_fn=panel.bed --gvcf\n\n# Benchmark against GIAB with stratification (the step that reveals ONT indel errors)\nhap.py giab_truth.vcf.gz clair3_out\u002Fmerge_output.vcf.gz \\\n  -f giab_confident.bed -r ref.fa --engine=vcfeval \\\n  --stratification giab_stratifications.tsv -o bench\u002Fhg002\n```\n\n## Per-Method Failure Modes\n\n### Silent model mismatch\n**Trigger:** `--model_path` pointing at a model that does not match the basecaller chemistry\u002Ftier\u002Fversion. **Mechanism:** no auto-detection; the wrong network runs. **Symptom:** no error, lower F1 (indels most). **Fix:** derive the model from the basecaller string; verify the folder exists (rerio for the full set); for v2 ensure `.pt` models.\n\n### Clair3 found nothing on a non-human reference\n**Trigger:** bacterial genome or draft assembly without chr1-22,X,Y names. **Mechanism:** Clair3 calls only standard human contigs by default. **Symptom:** near-empty VCF. **Fix:** `--include_all_ctgs`.\n\n### Global F1 looks great, clinical genes are wrong\n**Trigger:** reporting only whole-genome F1. **Mechanism:** ONT indel errors concentrate in homopolymer\u002FSTR\u002Flow-complexity strata. **Symptom:** ~99.5% global indel F1 but much lower in LowComplexity. **Fix:** stratify with GIAB BEDs (Dwarshuis 2024); use CMRG for medically relevant genes.\n\n### Treating Clair3 as a somatic caller\n**Trigger:** lowering `--snp_min_af`\u002F`--indel_min_af` to catch low-VAF variants. **Mechanism:** germline model expects ~0.5\u002F1.0 allele fractions, is not VAF-aware. **Symptom:** germline-model false positives at low AF, missed true somatic. **Fix:** ClairS (paired) \u002F ClairS-TO (tumor-only).\n\n### v1 model with v2 Clair3\n**Trigger:** an old TensorFlow model dir with Clair3 v2. **Mechanism:** v2 needs PyTorch `.pt` models. **Symptom:** model load failure. **Fix:** use `pileup.pt`\u002F`full_alignment.pt` models (Converted Rerio).\n\n## Quantitative Thresholds\n\n| Threshold | Source | Rationale |\n|-----------|--------|-----------|\n| Recommended depth ~20-60x | Clair3 guidance | sensitivity (hets, indels) falls off below ~20x; `--min_coverage` default 2 is a floor, not a recommendation |\n| Phasing buys ~6% indel F1 | Zheng 2022 | haplotagged reads disambiguate indel alleles in repeats |\n| ONT R10.4.1 sup: SNP F1 ~99.99%, indel F1 ~99.5% | GIAB benchmarks | indel residual lives in homopolymer\u002FSTR strata |\n| `--var_pct_full` 0.3 (default) | Clair3 README | fraction of low-quality pileup calls re-run by full-alignment; raise for recall, slower |\n| Stratify with GIAB \u002F CMRG | Dwarshuis 2024 | global F1 hides the ONT indel problem |\n\n## Common Errors\n\n| Error \u002F symptom | Cause | Solution |\n|-----------------|-------|----------|\n| Empty\u002Fnear-empty VCF on non-human ref | default calls only chr1-22,X,Y | `--include_all_ctgs` |\n| Model fails to load | v1 TF model with v2 Clair3 | use `.pt` (PyTorch) models |\n| `--model_path ...\u002Fmodels\u002Font` not found | no generic `ont`\u002F`hifi` model | point at a specific model subfolder |\n| Worse-than-expected indels | wrong-version or wrong-tier model | match the basecaller model exactly |\n| \"joint genotyping\" gave odd merges | `bcftools merge` on gVCFs is not joint calling | use GLnexus |\n| Looking for somatic\u002Flow-VAF variants | germline caller | use ClairS \u002F ClairS-TO |\n\n## References\n\n- Zheng Z, Li S, Su J, Leung AWS, Lam TW, Luo R. 2022. Symphonizing pileup and full-alignment for deep learning-based long-read variant calling (Clair3). *Nat Comput Sci* 2:797-803.\n- Zheng Z, He M, Yu X, et al. 2026. Accelerated long-read variant calling with Clair3 for whole-genome sequencing. *Bioinformatics* (advance access) btag181.\n- Kolesnikov A, Cook D, Nattestad M, et al. 2024. Local read haplotagging enables accurate long-read small variant calling. *Nat Commun* 15:5907.\n- Dwarshuis N, Kalra D, McDaniel J, et al. 2024. The GIAB genomic stratifications resource for human reference genomes. *Nat Commun* 15:9029.\n- Lin JH, Chen LC, Yu SC, Huang YT. 2022. LongPhase: an ultra-fast chromosome-scale phasing algorithm for small and large variants. *Bioinformatics* 38(7):1816-1822.\n- Chen L, Zheng Z, Su J, et al. 2025. ClairS-TO: a deep-learning method for long-read tumor-only somatic small variant calling. *Nat Commun* 16:9630.\n\n## Related Skills\n\n- basecalling - The basecaller model+version the Clair3 model must match\n- long-read-alignment - Produces the BAM (keep `--MD`; use minimap2 >=2.28)\n- haplotype-phasing - whatshap\u002Flongphase phasing and haplotagging Clair3 uses internally\n- medaka-polishing - ONT consensus; medaka diploid variant calling is deprecated in favor of Clair3\n- structural-variants - SVs are out of Clair3's scope (Sniffles2\u002FcuteSV)\n- variant-calling\u002Fdeepvariant - DeepVariant native ONT\u002FHiFi models (neck-and-neck on R10)\n- variant-calling\u002Fvcf-statistics - Summarize\u002Ffilter the VCF Clair3 emits\n- clinical-databases\u002Fvariant-prioritization - Prioritize the called variants\n",{"name":5,"source":24,"license":15,"tool_type":25,"description":7,"imported_at":26,"source_path":27,"primary_tool":28,"source_repository":29},"https:\u002F\u002Fgithub.com\u002FGPTomics\u002FbioSkills","cli","2026-09-20","long-read-sequencing\u002Fclair3-variants\u002FSKILL.md","Clair3","GPTomics\u002FbioSkills","main","2d1750717747097b602acf30bd09a7cbd4b4c6ea","awesome_skills\u002Fbiology\u002Fbio-long-read-sequencing-clair3-variants","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous","2026-11-09 11:45:36",{"source":8,"skillName":5,"commands":37,"prompt":47,"zipAvailable":48},[38,41,44],{"tool":39,"command":40},"npx","npx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-long-read-sequencing-clair3-variants",{"tool":42,"command":43},"bunx","bunx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-long-read-sequencing-clair3-variants",{"tool":45,"command":46},"pnpm","pnpm dlx skills add Mr-Nobody-Anonymous\u002FAll-skills --skill bio-long-read-sequencing-clair3-variants","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-long-read-sequencing-clair3-variants\u002FSKILL.md",true,[50,56],{"path":51,"resourceType":52,"resourceTypeDesc":53,"sizeBytes":54,"sourceHtmlUrl":55},"examples\u002Fclair3_workflow.sh","SCRIPT","脚本",1524,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-long-read-sequencing-clair3-variants\u002Fexamples\u002Fclair3_workflow.sh",{"path":57,"resourceType":58,"resourceTypeDesc":59,"sizeBytes":60,"sourceHtmlUrl":61},"usage-guide.md","REFERENCE","参考文档",4589,"https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fbiology\u002Fbio-long-read-sequencing-clair3-variants\u002Fusage-guide.md",[63,71,78,85,92,99],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":70},"01M7FFFZPC2025JQQHPHDX154J","dei-assessment","Assess diversity, equity, and inclusion with metrics. TRIGGERS - Use when user needs help with dei-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-assessment\u002FSKILL.md",[],[],"2026-11-16 10:05:34",{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":70},"01M7FFFWN8H4WA7A6Y398QKB6J","dei-action-plan","Create DEI action plans with goals and initiatives. TRIGGERS - Use when user needs help with dei-action-plan related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fdei-action-plan\u002FSKILL.md",[],[],{"skillNo":79,"name":80,"nameZh":6,"description":81,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":70},"01M7FFFSHTM0QZRVZ2NRJTJ0PY","culture-fit-assessment","Design culture add assessment processes. TRIGGERS - Use when user needs help with culture-fit-assessment related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fculture-fit-assessment\u002FSKILL.md",[],[],{"skillNo":86,"name":87,"nameZh":6,"description":88,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":89,"categories":90,"keywords":91,"publishedAt":70},"01M7FFFMA12RKEQA38KHJSXXPS","cowork-to-code-bridge","Use an already-installed, independently verified cowork-to-code bridge to run narrowly approved actions on the user's own macOS, Linux, or WSL2 machine through a local file queue.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcowork-to-code-bridge\u002FSKILL.md",[],[],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":98,"publishedAt":70},"01M7FFFH1F6M8YRGW5V487GH1K","counter-offer-response","Draft counter-offer responses for retention. TRIGGERS - Use when user needs help with counter-offer-response related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcounter-offer-response\u002FSKILL.md",[],[],{"skillNo":100,"name":101,"nameZh":6,"description":102,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":103,"categories":104,"keywords":105,"publishedAt":70},"01M7FFFDGEHV8Y7J9N40HZXW52","continuous-feedback-system","Design continuous feedback systems. TRIGGERS - Use when user needs help with continuous-feedback-system related tasks.","https:\u002F\u002Fgithub.com\u002FMr-Nobody-Anonymous\u002FAll-skills\u002Fblob\u002Fmain\u002Fawesome_skills\u002Fhr-people\u002Fcontinuous-feedback-system\u002FSKILL.md",[],[]]