菜单
Germline and somatic short-variant calling pipeline for Illumina short-read data. Use when the user mentions BWA, BWA-MEM2, GATK, HaplotypeCaller, Mutect2, VCF, GVCF, VQSR, variant calling, germline, somatic, SNV, or indel calling from FASTQ/BAM.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Reasoning skill for healthcare revenue cycle denial management. Use when the user needs to appeal a claim denial, draft a payer-ready denial appeal letter, analyze a denial reason code, or determine whether a denied claim should be overturned, pended, or accepted. Triggers include "appeal a claim denial", "draft denial appeal", "write appeal letter", "CO-50 denial", "medical necessity denial", "CARC code appeal", "RARC code", "overturn denial", "prior auth denial", "timely filing denial", "denial workup", "payer denial response", "claim denial overturn", "CO-96 denial", "CO-97 denial", "denial letter", "denial management", "remittance advice denial", "835 denial", "authorization required denial", "ERISA denial", "Medicaid managed care denial", "visit limit denial", "benefit limit denial", "inpatient denial", "observation downgrade".
Reasoning skill for designing AWS GenAI and ML architectures for healthcare and life sciences workloads. Use when the user asks to choose between SageMaker and Bedrock, design a RAG system over medical literature, architect clinical NLP or medical imaging inference, plan genomics or drug discovery pipelines on AWS, address HIPAA/PHI compliance in ML systems, design MLOps for regulated clinical models, or optimize cost for HCLS ML workloads. Triggers include "AWS architecture", "SageMaker vs Bedrock", "HIPAA ML", "clinical RAG", "medical imaging inference", "genomics on AWS", "PHI training", "MLOps healthcare", "Bedrock guardrails", "HealthLake", "HCLS cloud architecture", "BAA compliance", "SageMaker endpoint", "Bedrock knowledge base", "clinical NLP on AWS", "FDA SaMD on AWS".
Reason about biomarker discovery and validation in HCLS — classifying biomarker intent, choosing feature-selection and cross-validation strategies, avoiding leakage, and planning external replication. Use when the user asks to discover, develop, or validate a biomarker; select features from high-dimensional omics or clinical data; design a validation study; choose evaluation metrics; justify sample size; combine multi-omics signals; or assess clinical utility. Triggers include "discover a biomarker", "validate biomarker", "prognostic vs predictive", "feature selection", "LASSO vs elastic net", "nested cross-validation", "data leakage", "C-index", "time-dependent AUC", "decision curve analysis", "external validation cohort", "events per variable", "optimism-corrected", "multi-omics integration", "clinical utility of a biomarker", "is this biomarker ready".
Reason about CDISC SDTM and ADaM implementation for regulatory submissions. Use when the user asks about SDTM domain mapping, ADaM dataset design, controlled terminology versioning, define.xml completeness, FDA or PMDA submission requirements, query prioritization by clinical impact, SUPPQUAL usage, or CDISC compliance review. Triggers include "SDTM mapping", "ADaM dataset", "CDISC compliance", "controlled terminology", "define.xml", "FDA submission data", "PMDA submission", "SDTM domain", "ADSL", "ADAE", "ADLB", "BDS structure", "SUPPQUAL", "RELREC", "value-level metadata", "CDISC CT", "regulatory submission data standards", "eCTD datasets", "SDTM 3.3", "ADaM 1.1", "query prioritization", "clinical data review".
Generate code to assign cell type labels to single-cell RNA-seq clusters using CellTypist, SingleR, marker-based annotation, or reference label transfer (scANVI/ingest). Triggers on requests to "annotate cell types", "label clusters", "run CellTypist", "SingleR annotation", "marker gene dotplot", "transfer labels from reference atlas", "cell identity", "automated annotation", "reference mapping", "scANVI label transfer", "canonical markers", "immune cell types", "hierarchical annotation", "majority voting CellTypist", "over-clustering annotation".
Cheminformatics pipeline for small-molecule property calculation, filtering, and similarity analysis using RDKit. Use when the user asks to compute molecular descriptors, filter compounds by Lipinski or Veber rules, detect PAINS, calculate fingerprint similarity, run matched molecular pair analysis, generate ADMET descriptors, or process SMILES. Triggers include "RDKit", "molecular descriptors", "Lipinski", "rule of five", "Veber", "PAINS", "pan-assay interference", "Morgan fingerprint", "Tanimoto", "fingerprint similarity", "matched molecular pair", "MMP", "mmpdb", "ADMET", "druglikeness", "SMILES", "cheminformatics", "compound filtering", "chemical similarity".
| name | variant-calling |
| description | Germline and somatic short-variant calling pipeline for Illumina short-read data. Use when the user mentions BWA, BWA-MEM2, GATK, HaplotypeCaller, Mutect2, VCF, GVCF, VQSR, variant calling, germline, somatic, SNV, or indel calling from FASTQ/BAM. |
| usage | Invoke when running germline or somatic variant calling with BWA-MEM2, GATK4 HaplotypeCaller, Mutect2, or VQSR. |
| version | 1.0.0 |
| validated_against | {"date":"2025-01-15T00:00:00.000Z","packages":{"gatk":"4.5","bwa-mem2":"2.2.1","samtools":"1.19"}} |
| tags | ["skill","category:pipeline","genomics","variant-calling","gatk","bwa","hcls"] |
Adds decision logic for VQSR vs hard filters, WES vs WGS parameter differences, and GATK4-specific gotchas that LLMs frequently get wrong (filter thresholds, GVCF requirements, reference consistency).
Germline calling strategy:
├── Single sample, no future joint calling → HaplotypeCaller direct VCF
└── Cohort (≥2 samples) → HaplotypeCaller -ERC GVCF → GenomicsDBImport → GenotypeGVCFs
Filtering strategy:
├── ≥30 WGS samples OR ≥30 WES exomes → VQSR
└── <30 samples OR single sample → Hard filters
Somatic calling:
├── Tumor + matched normal → Mutect2 with both BAMs
└── Tumor-only → Mutect2 + PoN (MANDATORY) + gnomAD germline resource
WES vs WGS parameter differences:
| Step | WGS | WES |
|---|---|---|
| BQSR / HaplotypeCaller | No -L | -L capture.bed -ip 100 |
| Known sites for BQSR | Genome-wide | Genome-wide (do NOT subset) |
| Filtering | VQSR (≥30 samples) | Hard filters unless ≥30 exomes |
| Joint genotyping intervals | Per-chromosome | -L capture.bed |
Hard filter thresholds (GATK Best Practices):
| Filter | SNP threshold | Indel threshold |
|---|---|---|
| QD | < 2.0 | < 2.0 |
| FS | > 60.0 | > 200.0 |
| MQ | < 40.0 | — |
| MQRankSum | < -12.5 | — |
| ReadPosRankSum | < -8.0 | < -20.0 |
| SOR | > 3.0 | > 10.0 |
VQSR truth sensitivity levels:
--max-gaussians 4 (fewer training variants than SNPs)VQSR annotations: -an QD -an FS -an MQ -an MQRankSum -an ReadPosRankSum -an SOR
Mutect2 essentials:
--germline-resource af-only-gnomad.hg38.vcf.gz--panel-of-normals pon.vcf.gz (essential for tumor-only)--f1r2-tar-gz → LearnReadOrientationModel (FFPE/OxoG artifacts)GetPileupSummaries + CalculateContamination before FilterMutectCallsWrong: Aligning without proper @RG headers (missing ID, SM, PL, LB)
Right: Always specify at alignment: -R '@RG\tID:x\tSM:x\tPL:ILLUMINA\tLB:x'
Why: GATK refuses to run or silently merges samples when SM tags are wrong
Wrong: Running HaplotypeCaller without -ERC GVCF for cohort analysis
Right: Always produce GVCFs when joint genotyping will be performed
Why: Regular VCFs cannot be joint-genotyped; must re-call from BAM
Wrong: Reusing SNP hard-filter thresholds for indels (e.g., FS > 60 for indels)
Right: Use FS > 200 for indels, FS > 60 for SNPs
Why: Indels tolerate higher strand bias; SNP thresholds over-filter real indels
Wrong: Applying VQSR to <30 samples Right: Use hard filters for small cohorts Why: VQSR needs many variants to train its Gaussian mixture model
Wrong: Subsetting --known-sites VCFs to WES capture BED for BQSR
Right: Use genome-wide known-sites; only restrict analysis intervals via -L
Why: BQSR needs genome-wide known sites to model base quality errors
Wrong: Skipping samtools index between GATK steps
Right: Index after every BAM-producing step
Why: GATK requires BAM indices; missing them causes immediate failure
Wrong: Mixing reference files from different genome builds Right: Ensure ref.dict, ref.fa.fai, BWA index, and all VCFs match the same build Why: Mismatched dictionaries cause silent failures or cryptic errors
Wrong: Running Mutect2 tumor-only without a panel of normals
Right: Always provide --panel-of-normals for tumor-only calling
Why: Without PoN, recurrent sequencing artifacts are called as somatic mutations