Orchestrates neoantigen discovery from somatic variants to ranked vaccine candidates, chaining HLA typing (OptiType/arcasHLA + LOHHLA), VEP annotation (Wildtype+Frameshift plugins) + expression/readcount annotation, proximal-variant phasing, pVACseq MHC-I/II binding, CCF/clonality, and immunogenicity/quality ranking. Use when recognizing that binding is single-digit PPV and the critical steps are downstream (full-resolution HLA + LOH gating, proximal-variant phasing, clonality from purity+CN not raw VAF, expression), sequencing normalize+annotate -> phase -> HLA -> binding -> quality in the defensible order, dropping candidates on LOH-lost alleles, supplying --phased-proximal-variants-vcf so the mutant peptide is real, or ranking WITHIN patient rather than a fixed IC50 threshold. Hands mechanism to the immunoinformatics component skills; not a re-teach of any single step.
Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.
Quelldateien prüfen
Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.
Mit Codex oder Claude installieren Kopieren Sie diesen Prompt, fügen Sie ihn in Codex, Claude oder einen anderen Assistant ein und lassen Sie die Skill-Seite prüfen und installieren.
Ein direkter Befehl überspringt den Prüf-Prompt. Prüfen Sie die Quelle, bevor Sie ihn ausführen.
Orchestrates neoantigen discovery from somatic variants to ranked vaccine candidates, chaining HLA typing (OptiType/arcasHLA + LOHHLA), VEP annotation (Wildtype+Frameshift plugins) + expression/readcount annotation, proximal-variant phasing, pVACseq MHC-I/II binding, CCF/clonality, and immunogenicity/quality ranking. Use when recognizing that binding is single-digit PPV and the critical steps are downstream (full-resolution HLA + LOH gating, proximal-variant phasing, clonality from purity+CN not raw VAF, expression), sequencing normalize+annotate -> phase -> HLA -> binding -> quality in the defensible order, dropping candidates on LOH-lost alleles, supplying --phased-proximal-variants-vcf so the mutant peptide is real, or ranking WITHIN patient rather than a fixed IC50 threshold. Hands mechanism to the immunoinformatics component skills; not a re-teach of any single step.
[{"after_hla":"HLA types resolved to 4-digit, coverage adequate"},{"after_binding":"Predictions for ALL alleles (LOH-lost alleles dropped); ranked within patient, not hard-thresholded across patients"},{"after_neoantigen":"Expressed (RNA-confirmed); clonality via CCF from purity+CN, not raw VAF"},{"after_scoring":"Top candidates are a tier-1 hypothesis list for MS + T-cell validation"}]
Before using code patterns, verify installed versions match. If versions differ:
Python: pip show <package> then help(module.function) to check signatures
CLI: <tool> --version then <tool> --help to confirm flags
If code throws ImportError, AttributeError, or TypeError, introspect the installed
package and adapt the example to match the actual API rather than retrying.
Neoantigen Pipeline
"Predict neoantigens from my tumor sequencing data" -> Orchestrate HLA typing (OptiType), somatic variant calling, pVACtools neoantigen prediction, MHC binding scoring, and immunogenicity-based candidate ranking for personalized cancer immunotherapy.
Complete workflow from somatic variants to ranked neoantigen vaccine candidates for personalized cancer immunotherapy.
Key Judgment -- binding is the easy part; PPV lives downstream
A binding-only pipeline has single-digit-percent positive predictive value (TESLA; Wells 2020 Cell 183:818). The critical steps are downstream of binding: correct full-resolution HLA typing (wrong allele = confident garbage), HLA loss-of-heterozygosity (run LOHHLA and DROP candidates on a lost allele; it invalidates predictions silently), proximal-variant phasing (supply --phased-proximal-variants-vcf or the mutant peptide is wrong), cancer cell fraction for clonality (clonal beats subclonal; use purity + copy number, not raw VAF), expression, and quality features (agretopicity, foreignness). Treat the ranked output as a tier-1 hypothesis list for immunopeptidomics MS and functional T-cell validation, not a final answer. Add MHC class II (CD4) neoantigens for vaccine help (see immunoinformatics/mhc-class-ii-prediction). Note on DAI below: agretopicity is most often the WT/MT binding ratio; whichever form is used, an anchor-position mutation inflates it without changing the TCR-facing surface, and a barely-presented WT makes it unstable; pair it with anchor evaluation.
Made-once commitments
Commitment
Consequence inherited downstream
HLA typing at full 4-digit resolution (class I + II)
A wrong allele is confident garbage; every binding prediction inherits it; reconcile DNA vs RNA calls
Order-trap 1 - normalize + annotate with the RIGHT plugins BEFORE pVACseq. pVACseq needs the Wildtype plugin (matched WT peptide -> agretopicity) and the Frameshift plugin (novel ORF); Frameshift REPLACED the legacy Downstream in pVACtools 2.0+. Normalize before annotate.
Order-trap 2 - PHASE proximal variants BEFORE translating the mutant peptide. THE review-sinker: editing variants independently yields a peptide the patient never makes. Merge somatic+germline, phase (WhatsHap/GATK), supply --phased-proximal-variants-vcf.
Order-trap 3 - HLA typing (+ LOHHLA) BEFORE binding. Binding is per-allele; a wrong or lost allele makes every downstream prediction garbage. Drop LOH-lost alleles before ranking.
Order-trap 4 - CCF/clonality from purity+copy-number BEFORE calling something subclonal. Low purity makes clonal look subclonal; correct VAF to cancer-cell fraction (copy-number/allele-specific-copy-number). Clonal beats subclonal.
Order-trap 5 - rank WITHIN patient; do NOT hard-threshold IC50 across patients. Immunogenicity scores are relative.
Workflow Overview
Somatic VCF (annotated) + Tumor RNA-seq (optional)
|
v
[1. HLA Typing] --> arcasHLA / OptiType (if types not provided)
|
v
[2. MHC Binding Prediction] --> MHCflurry / NetMHCpan
|
v
[3. Neoantigen Calling] --> pVACseq
|
v
[4. Immunogenicity Scoring] --> Multi-factor ranking
|
v
Ranked Vaccine Candidates (TSV + visualizations)
import json
withopen('hla_output/tumor.genotype.json') as f:
hla_data = json.load(f)
hla_alleles = []
for gene, alleles in hla_data.items():
for allele in alleles:
# arcasHLA emits 3-field alleles (A*01:01:01); pVACseq/IEDB validate 2-field (HLA-A*01:01)
hla_alleles.append('HLA-' + ':'.join(allele.split(':')[:2]))
# Format for pVACseq: HLA-A*02:01,HLA-A*24:02,HLA-B*07:02,...
hla_string = ','.join(hla_alleles)
print(f'HLA alleles: {hla_string}')
Step 2: VCF Annotation with VEP
pVACseq requires VEP-annotated VCF with specific fields:
# Annotate somatic VCF
vep --input_file somatic.vcf \
--output_file somatic.vep.vcf \
--format vcf --vcf --symbol --terms SO \
--plugin Frameshift --plugin Wildtype \
--offline --cache \
--pick --fork 4
# Add expression data (optional but recommended)# Positionals: <vcf> <expression_file> {kallisto,stringtie,cufflinks,custom} {gene,transcript}
vcf-expression-annotator somatic.vep.vcf \
expression.tsv custom gene \
-s tumor_sample --id-column gene_id --expression-column tpm \
-o somatic.vep.expression.vcf
# PHASE proximal variants (the review-sinker). Merge somatic + germline, phase with WhatsHap,# and pass the result to pVACseq via --phased-proximal-variants-vcf so a second variant in the# same codon-window (esp. a germline SNP in cis) yields the peptide the tumor ACTUALLY makes.
whatshap phase -o phased.vcf.gz --reference reference.fa somatic_plus_germline.vcf.gz tumor.bam
tabix -p vcf phased.vcf.gz
Step 3: Run pVACseq (Ensembl VEP 111+)
# Basic run with MHC Class I
pvacseq run \
somatic.vep.vcf \
tumor_sample \
"HLA-A*02:01,HLA-A*24:02,HLA-B*07:02,HLA-B*44:02,HLA-C*07:02,HLA-C*05:01" \
MHCflurry MHCnuggetsI NetMHCpan \
pvacseq_output/ \
-e1 8,9,10,11 \
--iedb-install-directory /path/to/iedb \
-t 8
# With expression filtering
pvacseq run \
somatic.vep.expression.vcf \
tumor_sample \
"HLA-A*02:01,HLA-A*24:02,HLA-B*07:02,HLA-B*44:02" \
MHCflurry NetMHCpan \
pvacseq_output/ \
-e1 8,9,10,11 \
--phased-proximal-variants-vcf phased.vcf.gz \
--tumor-purity 0.7 \
--tdna-vaf 0.1 \
--expn-val 1 \
-t 8
Drop candidates on HLA-LOH-lost alleles (run LOHHLA/DASH) BEFORE ranking, and correct clonality to cancer-cell fraction (CCF from purity + copy number, not raw VAF; see copy-number/allele-specific-copy-number). The raw-VAF filter below is a coarse proxy.
MT binds better than WT (>1); higher = more tumor-specific
Common Errors
Symptom
Cause
Fix
Peptides the tumor never makes
Proximal variants edited independently (unphased)
--phased-proximal-variants-vcf (WhatsHap/GATK); include germline in cis
Frameshift ORFs lost / no agretopicity
Wrong/legacy VEP plugin (Downstream instead of Frameshift; missing Wildtype)
pvacseq install_vep_plugin; run --plugin Wildtype --plugin Frameshift
Confident but invalid predictions
HLA allele wrong or on a LOH-lost haplotype
Full 4-digit typing + LOHHLA drop before ranking
--expn-val/VAF filters silently pass everything
Expression/readcounts not annotated into the VCF
vcf-expression-annotator + vcf-readcount-annotator before pVACseq
Clonal candidate mis-tiered subclonal
Raw VAF used as clonality on a low-purity tumor
CCF from purity + copy number (copy-number/allele-specific-copy-number)
Candidates mis-ranked across patients
Fixed IC50 threshold applied cross-patient
Rank WITHIN patient (immunoinformatics/immunogenicity-scoring)
No neoantigens found
Low mutation burden
Lower IC50 threshold to 1000nM; check TMB/MSI first
References
Hundal J, Kiwala S, McMichael J, et al (2020) pVACtools: a computational toolkit to identify and visualize cancer neoantigens. Cancer Immunology Research 8:409-420. DOI 10.1158/2326-6066.CIR-19-0401.
Wells DK, van Buuren MM, Dang KK, et al (2020) Key parameters of tumor epitope immunogenicity revealed through a consortium approach improve neoantigen prediction (TESLA). Cell 183:818-834. DOI 10.1016/j.cell.2020.09.015. (single-digit PPV of binding-only.)
McGranahan N, Rosenthal R, Hiley CT, et al (2017) Allele-specific HLA loss and immune escape in lung cancer evolution. Cell 171:1259-1271. DOI 10.1016/j.cell.2017.10.001. (LOHHLA.)
Wood MA, Nguyen A, Struck AJ, et al (2020) neoepiscope improves neoepitope prediction with multivariant phasing. Bioinformatics 36:713-720. DOI 10.1093/bioinformatics/btz653. (phasing matters.)
Output Files
File
Description
*.filtered.tsv
pVACseq filtered neoantigens
*.all_epitopes.tsv
All predicted epitopes
top_neoantigen_candidates.tsv
Ranked vaccine candidates
neoantigen_summary.pdf
Visualization figures
Related Skills
immunoinformatics/mhc-binding-prediction - MHCflurry parameters; BA vs EL, %Rank vs nM, abundance bias
immunoinformatics/mhc-class-ii-prediction - class II (CD4) neoantigens for vaccine help
clinical-databases/hla-typing - HLA typing (T1K is the 2024-2026 all-rounder; OptiType for class I; arcasHLA for RNA-seq); check HLA-LOH via LOHHLA / DASH which abolishes neoantigen presentation in ~17% pan-cancer (~30%+ HNSCC / NSCLC / cervical)
