vaxrank

Vaxrank is the neoantigen ranking component of the OpenVax pipeline for designing personalized cancer vaccines. Given either (a) a patient's somatic mutations + tumor RNA-seq + HLA type, or (b) a pre-computed neoepitope report from LENS or pVACseq, Vaxrank selects and ranks the mutant antigens most likely to elicit a T-cell response and emits them as the vaccine type(s) the user requests — peptide pools, mRNA constructs, or analysis reports for review.

Contents

• Quick Start
• Overview
• Vaccine designs
• Vaccine types and output modes
• Clinical Use
• Installation
• Configuration
• MHC Binding Predictors
• How It Works
• Papers & Citations
• Dependencies
• Legacy flags
• Development

Quick Start

Full pipeline from a tumor VCF + RNA-seq BAM, emitting ranked vaccine peptides as text, HTML, and PDF reports:

vaxrank \
    --vcf tests/data/b16.f10/b16.vcf \
    --bam tests/data/b16.f10/b16.combined.bam \
    --mhc-predictor netmhc \
    --mhc-alleles H2-Kb,H2-Db \
    --output-ascii-report vaccine-peptides.txt \
    --output-pdf-report vaccine-peptides.pdf

Required inputs:

• --vcf — somatic variants (VCF from any variant caller)
• --bam — tumor RNA-seq alignments (used by Isovar to assemble mutant transcripts)
• --mhc-alleles — patient HLA alleles (e.g. HLA-A*02:01,HLA-B*07:02)
• --mhc-predictor — which MHC binding predictor to use (see MHC Binding Predictors)

Drive design from a pre-computed neoepitope report (LENS or pVACseq) when upstream MHC prediction has already been done:

vaxrank --input-lens patient.lens.tsv \
        --vaccine-type mrna --output-dir mrna_out/ \
        --ensembl-release 102

Emit both peptide and mRNA constructs in one run — outputs land in per-modality subdirs:

vaxrank --vcf v.vcf --bam r.bam \
        --vaccine-type peptide mrna --output-dir vaccines/
# → vaccines/peptide/{vaccine.fasta, manifest.json, order_form.csv}
# → vaccines/mrna/{cds.fasta, no_polyA.fasta, full.fasta, manifest.json, layers.csv}

Overview

Personalized cancer vaccines (also called neoantigen vaccines) work by training the immune system to recognise peptides that arise from somatic mutations unique to a patient's tumor. Designing such a vaccine requires a computational pipeline that bridges raw sequencing data and the peptide synthesiser:

1. Variant calling — Whole-exome or whole-genome sequencing of the tumor and matched normal identifies somatic mutations. This is typically done with tools such as MuTect or Strelka, upstream of Vaxrank.
2. Mutant transcript assembly — Tumor RNA-seq reads overlapping each mutation are assembled by Isovar to determine the true mutant protein sequence. This step phases nearby germline variants and captures any mutation-associated splicing differences, producing a more accurate reading frame than DNA-only prediction.
3. MHC binding prediction — Candidate epitopes (short peptide subsequences spanning the mutation) are scored for predicted binding to the patient's HLA class I molecules using mhctools, which wraps predictors such as MHCflurry, NetMHCpan, and BigMHC.
4. Vaccine peptide selection — Vaxrank assembles longer synthetic long peptides (SLPs, typically 25-mers) around the mutation, scores them by the number and strength of their predicted MHC-binding epitopes, filters out peptides that appear in the reference proteome, annotates known cancer hotspot mutations, and ranks candidates by a combined immunogenicity and manufacturability score.
5. Vaccine-type dispatch — the ranked candidates are written out as one or more of the vaccine types selected via --vaccine-type: a peptide pool ready for synthesis, an mRNA construct ready for IVT, or both. Analysis reports are emitted independently. Steps 1-3 are skipped when an external neoepitope report is supplied via --input-lens or --input-pvacseq; the ranking and dispatch steps are identical.

Vaccine designs

Vaxrank's vaccine design space is two orthogonal axes (shared across vaccine types) plus the type itself:

Combined, the matrix yields 8 distinct designs — 4 per vaccine type:

A third knob, --epitopes-per-antigen, controls how many top MHC ligands to take per ranked vaccine peptide when content is minimal_epitope. The default 1 is the "single top ligand" semantics; >1 packs multiple top ligands from the same variant as separate antigens.

Peptide designs

SLP (default). Mutation-spanning long peptide, one antigen per construct — the PGV-001 canonical design.

vaxrank --vcf v.vcf --bam r.bam --output-dir vaccine_out/
# → vaccine_out/{vaccine.fasta, manifest.json, order_form.csv}

Multi-epitope concatenated peptide. Several mutation-spanning antigens linked into one longer peptide. Use --peptide-linker to pick the spacer; AAY is the proteasome-friendly default.

vaxrank --vcf v.vcf --bam r.bam \
        --output-dir vaccine_out/ \
        --peptide-antigens-per-construct 5 --peptide-linker AAY

Minimal-epitope peptide. A single short MHC ligand per construct — useful when minimum-length manufacturability matters more than flanking context.

vaxrank --vcf v.vcf --bam r.bam \
        --output-dir vaccine_out/ \
        --antigen-content minimal_epitope

mRNA designs

BioNTech FixVac / iNeST canonical. Multi-antigen mutation-spanning mRNA — the default for --vaccine-type mrna. Antigens are linked with (G4S)2 and emitted as CDS, no-polyA, and full (with polyA) FASTAs plus a structured manifest.

vaxrank --vcf v.vcf --bam r.bam --vaccine-type mrna --output-dir mrna_out/
# → mrna_out/{cds.fasta, no_polyA.fasta, full.fasta, manifest.json, layers.csv}

String-of-beads mRNA. Concatenated minimal-epitope antigens — short MHC ligands linked together rather than mutation-spanning windows.

vaxrank --vcf v.vcf --bam r.bam --vaccine-type mrna --output-dir out/ \
        --mrna-antigen-content minimal_epitope --mrna-antigens-per-construct 8 \
        --mrna-linker AAY

Top-N ligands per variant in a string-of-beads. Pack multiple top MHC ligands from each ranked vaccine peptide as separate antigens.

vaxrank --vcf v.vcf --bam r.bam --vaccine-type mrna --output-dir out/ \
        --mrna-antigen-content minimal_epitope \
        --mrna-epitopes-per-antigen 2 --mrna-antigens-per-construct 16

Both modalities in one run

Multi-valued --vaccine-type writes per-modality subdirs in --output-dir.

vaxrank --vcf v.vcf --bam r.bam --vaccine-type peptide mrna --output-dir vaccines/
# → vaccines/peptide/{vaccine.fasta, manifest.json, order_form.csv}
# → vaccines/mrna/{cds.fasta, no_polyA.fasta, full.fasta, manifest.json, layers.csv}

Vaccine types and output modes

Vaxrank always ranks. The vaccine-type writer fires only when both --vaccine-type and --output-dir are set. --vaccine-type is multi-valued (default peptide): pass one or more of peptide / mrna. Single-mode runs write canonical files directly in --output-dir; multi-mode runs scope into per-modality subdirs (DIR/peptide/, DIR/mrna/, …). Analysis reports use their own --output-* flags and are independent of the vaccine-type dispatch.

# Peptide pool (default vaccine type)
vaxrank --vcf v.vcf --bam r.bam --output-dir vaccine_out/

# mRNA construct
vaxrank --vcf v.vcf --bam r.bam --vaccine-type mrna --output-dir mrna_out/

# Both at once (per-modality subdirs in mixed_out/)
vaxrank --vcf v.vcf --bam r.bam --vaccine-type peptide mrna --output-dir mixed_out/

# Reports only (no vaccine constructs)
vaxrank --vcf v.vcf --bam r.bam --output-pdf-report report.pdf

# Drive vaccine design from a pre-computed LENS report
vaxrank --input-lens patient.lens.tsv --vaccine-type mrna \
        --output-dir mrna_out/ \
        --ensembl-release 102

# Full ASCII summary report from a LENS file (transcripts resolved)
vaxrank --input-lens patient.lens.tsv --output-ascii-report report.txt \
        --ensembl-release 102

Analysis reports

Per-variant tables of ranked vaccine peptide candidates, predicted epitopes, and manufacturability scores. Independent of vaccine-type dispatch — runs whenever any report flag is set.

Mutation and assembly figures

vaxrank-mutation-figure turns the Isovar table produced by --output-isovar-csv into publication-ready, white-background transcript and protein context figures. When the input includes the optional *_cdna_sequence and *_cdna_variant_{start,end} columns, each variant gets a 5-prime-to-3-prime nucleotide comparison in addition to the protein view. Both views separate the reference transcript, annotation-only mutant prediction, and RNA-assembled sequence so unsupported or assembly-refined claims remain explicit.

vaxrank --vcf variants.vcf --bam tumor-rna.bam \
  --mhc-predictor random --mhc-alleles HLA-A*02:01 \
  --output-isovar-csv patient-isovar.csv

vaxrank-mutation-figure patient-isovar.csv \
  --output-root figures \
  --variant DYNC1H1 \
  --format svg --format pdf --format png \
  --png-scale 3

Outputs are grouped under figures/YYYY-MM-DDTHHMMSSZ/. Every variant gets an SVG, PDF, high-resolution PNG (3600×2100 at the default 3× scale), and compact JSON record. The figure and JSON distinguish annotation gene/transcript provenance from RNA-assembly provenance. The run manifest records the input checksum, Vaxrank version, formats, raster dimensions, and file inventory. Omit --variant to render every row. Use --timestamp for reproducible builds or tests; an existing run is never overwritten.

For structural variants and fusion junctions, use a cross-platform evidence JSON rather than an Isovar table:

vaxrank-evidence-figure evidence.json \
  --output-root evidence-figures \
  --format svg --format pdf --format png

This view keeps DNA, short-read RNA, short-read assembly, and long-read RNA as separate evidence stages. It displays the junction nucleotide sequence and withholds a protein unless the input supplies an established coding sequence. See examples/osteosarc_evidence_figures/ for DNA-only, cross-platform, matched-sample long-read rescue, and unresolved long-read-only examples.

Neoepitope report

Per-(peptide, allele) report (XLSX/CSV). Default output of the LENS / pVACseq input path; also available on the full pipeline.

Peptide constructs

vaccine.fasta + manifest.json + order_form.csv written into --output-dir (or --output-dir/peptide/ in multi-mode). The peptide design comes from --antigen-content and --peptide-antigens-per-construct (see Vaccine designs above).

mRNA constructs

A directory containing three FASTAs (cds.fasta, no_polyA.fasta, full.fasta) plus manifest.json (per-element view) and layers.csv (long-format per-element table with AA + nt). Codon optimization uses DnaChisel; 2A self-cleaving peptides preserve their published codon usage automatically.

Flags are grouped by what they configure:

Construct anatomy

PolyA tail

Linker optimization

Per-junction MHC-aware linker swap minimizes predicted presentation of chimeric k-mers spanning antigen junctions.

Codon optimization

External-input mode

Drive vaccine design from a pre-computed neoepitope report instead of VCF + BAM. Same downstream dispatch — peptide and mRNA construct outputs work identically.

Manifest schema

The peptide and mRNA construct JSON manifests share a back-compat schema (modality, name, length, length_unit, antigen_names, components, manufacturability). The mRNA manifest additionally exposes cds, no_polya_nt, full_nt, per-antigen antigens (each with AA + nt), and a structured elements dict with one entry per layer (5' UTR, signal peptide, antigens, linkers per junction, MITD, stop codon, 3' UTR, polyA) — every layer carrying both AA (where applicable) and nt forms for direct inspection.

Shared linker library and grammar

Both vaccine types consume the same set of linker names so a single construct design can be ported between peptide and mRNA backbones.

Static entries:

Compositional grammar (parsed at lookup time):

Repeat counts are capped at 100. 2A entries (codon-frozen, positional) are rejected in repeat forms — use the base linker once.

Every name resolves through vaccine_library.get_linker(name) and returns a Linker with primary-source citations attached. The default mRNA inter-antigen linker is (G4S)2 (BioNTech FixVac canonical, Sahin Nature 2017); the default peptide linker is G4S3. Per-junction MHC-aware linker swap (--mrna-optimize-linkers, on by default) considers G3S, G4S, (G3S)2, (G4S)2, AAA per junction and substitutes whichever minimizes predicted presentation of chimeric k-mers spanning the junction.

All sequences carry primary-source citations in vaxrank/vaccine_library.py.

Clinical Use

Vaxrank is the ranking engine behind the OpenVax neoantigen vaccine pipeline, which has been used in several clinical trials of personalized cancer vaccines at Mount Sinai:

• PGV001 (NCT02721043) — A phase I study of personalised neoantigen vaccines in patients with solid and haematologic malignancies. All 11 treated patients developed neoantigen-specific T-cell responses (Saxena et al., Cancer Discovery 2025).
• PGV001 + atezolizumab in urothelial cancer (NCT03359239) — A phase I trial combining PGV001 with checkpoint inhibition. The combination was safe and induced neoantigen-specific CD4+ and CD8+ T-cell responses in all evaluated patients (Saxena et al., Nature Cancer 2025).
• PGV001 + TTFields in newly diagnosed glioblastoma (NCT03223103) — A phase I trial combining PGV001 with tumor treating fields and standard-of-care temozolomide (paper in preparation).

The computational pipeline used in these trials is described in Kodysh & Rubinsteyn, Methods Mol. Biol. 2020.

Installation

pip install vaxrank

Requirements: Python 3.9+

Vaxrank uses PyEnsembl for reference genome annotation. Install an Ensembl release matching your reference genome:

# GRCh38
pyensembl install --release 113 --species human
# GRCh37 (legacy)
pyensembl install --release 75 --species human

PDF report generation uses wkhtmltopdf by default:

brew install --cask wkhtmltopdf

Alternatively, pass --pdf-backend=weasyprint to use WeasyPrint (experimental), which has no external binary dependency:

pip install weasyprint
# macOS also needs: brew install pango

On Apple Silicon, WeasyPrint loads Pango via dyld, which doesn't search Homebrew's /opt/homebrew/lib by default. Add this to your shell profile:

export DYLD_FALLBACK_LIBRARY_PATH="/opt/homebrew/lib:$DYLD_FALLBACK_LIBRARY_PATH"

(Intel macOS doesn't need this — Homebrew's /usr/local/lib is in dyld's default fallback path.)

Configuration

YAML config file

Common parameters can be stored in a YAML file to avoid repeating them on every run:

vaxrank --config my_config.yaml --vcf variants.vcf --bam tumor.bam

Example my_config.yaml:

epitopes:
  min_score: 0.00001                        # drop epitopes below this score
  scoring_mode: affinity                    # "affinity" or "percentile_rank"
  logistic_midpoint: 350.0                  # IC50 (nM) at which score = 0.5
  logistic_width: 150.0                     # steepness of logistic curve
  affinity_cutoff: 5000.0                   # IC50 >= this → score 0
  percentile_rank_cutoff: 10.0              # rank >= this → score 0 (percentile mode)
  top_epitopes_per_candidate: 1000          # 0 = keep all

vaccine_peptides:
  preferred_length: 25                      # target amino acids per vaccine peptide
  min_length: 25                            # minimum vaccine peptide length
  max_length: 25                            # maximum vaccine peptide length
  padding_around_mutation: null             # adaptive RNA context; explicit legacy override
  per_mutation: 1                           # peptides to keep per variant
  max_epitopes_per_candidate: 1000          # 0 = keep all
  score_fraction_of_best: 0.99              # drop candidates scoring < 99% of best
  manufacturability:                        # GRAVY = mean hydropathy
    max_c_terminal_hydropathy: 1.5          # max GRAVY of C-terminal 7-mer
    min_kmer_hydropathy: 0.0                # min max-7mer GRAVY (floor)
    max_kmer_hydropathy_low_priority: 1.5   # low-priority max-7mer GRAVY cap
    max_kmer_hydropathy_high_priority: 2.5  # high-priority max-7mer GRAVY cap

RNA reconstruction context

Vaxrank 3.14 requires Isovar ≥1.8.0. By default, the resolved vaccine peptide size K drives a 2*K-1 amino-acid context request: 29 for 15mers, 49 for 25mers, and 59 for 30mers. Actual RNA-supported context may be shorter. Isovar's balanced policy maximizes mutation-containing peptide windows among candidates retaining at least 85% of the best candidate's compatible read-name support. Independently, every retained cDNA base must meet a two-read-object coverage floor. The fraction is a configurable selection tolerance, not biological confidence or per-base depth; compatible reads need not span a whole vaccine peptide. No reference sequence is added to fill missing RNA context. This is a reconstruction policy, not a change to Vaxrank's MHC scoring or ranking model (see the PGV pipeline methods).

isovar:
  protein_sequence_length: null            # derive from peptide size
  protein_context_peptide_length: null     # use vaccine_peptides.preferred_length
  protein_sequence_preference: balanced    # or support / context
  min_protein_sequence_support_fraction: 0.85
  min_variant_sequence_coverage: 2

Each setting also has a same-named CLI flag with hyphens in place of underscores, or can be set using --config-value isovar.<name>=<value>. Explicit CLI values win over YAML, including values equal to the defaults and a coverage floor of zero (which explicitly disables that floor). An explicit protein length wins over legacy --padding-around-mutation / YAML padding; otherwise explicit padding requests K + 2*padding. With neither, Isovar derives the adaptive target. To reproduce historical Isovar extraction and support-first selection, pass --protein-sequence-length 20 --protein-sequence-preference support.

The target includes all placements for a centered single-residue mutation, not necessarily wider mutations; deletions require windows spanning their junction. Short outputs are not evidence of a full-length vaccine window. Vaxrank rejects candidate windows shorter than vaccine_peptides.min_length (25 by default); explicitly lower that minimum only when shorter peptides are part of the intended design. The RNA support thresholds remain unchanged. RNA settings are saved with run arguments and do not rerun reconstruction when rendering cached reports or ranking external pVACseq/LENS predictions. DNA-only fallback remains opt-in, with its existing separate padding default.

Custom filtering and scoring with the topiary DSL

For anything beyond the scalar logistic / percentile-rank defaults, set epitopes.filter_expr and/or epitopes.score_expr to a topiary DSL string. Both accept the full topiary 5.0 expression grammar (kind accessors like affinity / presentation, arithmetic, & / |, .logistic(...) / .clip(...) transforms, column(col_name) for raw DataFrame columns, etc.).

epitopes:
  # Drop rows wholesale before scoring
  filter_expr: "affinity <= 500 & affinity.rank <= 2.0"
  # Compute a per-(peptide, allele) score in [0, 1] (binder-quality score)
  score_expr:  "affinity.logistic_normalized(350, 150)"

When filter_expr is omitted, no rows are dropped up-front; the default score_expr is synthesized from the scalar fields above (binding_affinity_cutoff, logistic_midpoint, logistic_width, etc.) and masked so ic50 >= affinity_cutoff → 0, reproducing the pre-5.0 behavior byte-for-byte.

Use affinity.logistic_normalized(m, w) for a [0, 1] binder-quality score (the topiary 5.1+ primitive); the plain affinity.logistic(m, w) is the raw sigmoid and caps below 1 (≈0.912 at default m=350, w=150).

Invalid DSL strings are rejected at config load (not mid-pipeline), so typos in the YAML surface before any predictions run.

CLI overrides

CLI arguments override YAML values. You can also use --config-value to override individual keys without editing the file:

vaxrank --config my_config.yaml \
  --config-value vaccine_peptides.score_fraction_of_best=0.95 \
  --config-value epitopes.percentile_rank_cutoff=5.0

Use --config-text when the right-hand side should be kept as a raw string instead of being YAML-parsed.

Resolution order

Config values are resolved in order (later wins):

1. Compiled-in defaults (see vaxrank/config/defaults.py)
2. YAML config file (--config)
3. --config-value / --config-text overrides
4. Dedicated CLI flags (e.g. --vaccine-peptide-length)

Config reference

EpitopeConfig — epitope scoring and filtering

VaccineConfig — peptide assembly and manufacturability

The four *_hydropathy* fields control the manufacturability tie-breaking in vaccine peptide ranking. See VaccinePeptide.peptide_synthesis_difficulty_score_tuple for details on how each threshold is applied.

MHC Binding Predictors

Vaxrank integrates with MHC binding predictors via mhctools. Use --mhc-predictor <name> to select one:

How It Works

Upstream inputs

Vaxrank accepts two distinct input shapes, both producing the same ranked-vaccine-peptides intermediate:

Full pipeline (VCF + BAM): Vaxrank does not perform variant calling or read alignment itself. Those steps happen upstream, typically as part of a larger bioinformatics pipeline (e.g. neoantigen-vaccine-pipeline):

1. Tumor and matched-normal DNA are sequenced and aligned; a variant caller (MuTect, Strelka, etc.) produces a VCF of somatic mutations.
2. Tumor RNA is sequenced and aligned to produce a BAM file.
3. The patient's HLA class I alleles are typed (from sequencing data or clinical records).

Vaxrank takes these three inputs — the VCF, the tumor RNA BAM, and the HLA alleles — runs Isovar transcript assembly + MHC binding prediction

• ranking, and produces vaccine peptide candidates.

External-input mode (--input-lens or --input-pvacseq): when an upstream tool (e.g. LENS or pVACseq) has already produced a per-(peptide, allele) neoepitope report, Vaxrank skips Isovar + MHC prediction and consumes the report directly. The per-row pep_context (LENS) or Best Peptide / MT Epitope Seq (pVACseq) is used as the SLP-style antigen window. Downstream dispatch — reports + peptide constructs + mRNA constructs — is identical to the full pipeline. pVACseq parsing is delegated to topiary, so both all_epitopes.tsv and all_epitopes.aggregated.tsv are accepted.

LENS FUSION rows are also eligible for construct ranking. Vaxrank treats their caller-supplied pep_context as an assembled fusion antigen rather than fabricating a single-locus genomic variant. Fusion ID, both partner genes, breakpoints, transcript IDs, fusion type, RNA evidence fields, and sequence source are preserved in the report; ranked fusion antigens can be included in both peptide and mRNA constructs. This path imports LENS's peptide predictions. It does not infer a coding frame or translate raw breakpoints when the upstream report has not supplied an established protein context.

LENS SPLICE, CTA/SELF, and ERV rows are preserved in the neoepitope report but require an explicit opt-in before they enter construct ranking:

vaxrank --input-lens patient.lens.tsv \
  --include-antigen-source SPLICE \
  --include-antigen-source CTA/SELF \
  --include-antigen-source ERV

--exclude-antigen-source removes a default or configured category (for example, --exclude-antigen-source FUSION). The same exact selection can be set with vaccine_peptides.included_antigen_sources in YAML. Splice targets retain their genomic junction, description, coding sequence, and RNA evidence; CTA/self targets retain gene, transcript, normal-tissue, mTEC, and expression evidence; ERV targets retain the ORF identifier and all erv_* evidence. Only caller-designated peptide intervals are targetable because these LENS rows do not state an exact amino-acid junction offset. CTA/self and ERV opt-ins are recorded as review-required tumor-specificity overrides, not as proof of sequence novelty or absence from normal tissue.

For an upstream assembler that supplies the translated sequence and exact amino-acid junction, VaccineAntigen.from_fusion_sequence(...) records the junction as a zero-width target mask. predict_epitopes(..., antigen=antigen) then enumerates the predictor's peptide lengths and marks only peptides that contain residues from both fusion partners as targetable; no fictitious wild-type comparator is generated.

For validated results from isovar.reconstruct_fusion(...) or isovar fusion, use fusion_antigens_from_isovar(result, tumor_specificity=attestation, gene_name="DONOR::ACCEPTOR", species="Homo sapiens"). It returns an IsovarFusionAntigens containing every coding hypothesis in .antigens, the complete JSON-normalized result in .reconstruction, and the eligible subset in .admitted_antigens. No best transcript is selected. Nucleotide junctions are mapped exactly: a mixed codon is targetable, while peptides entirely inside an inserted sequence are not junction-spanning.

RNA validation is not tumor-specificity evidence: absent a separate TumorSpecificityAttestation, every hypothesis is held out. Ambiguous, unsupported and partial-CDS reconstructions remain held out even with an admission attestation; their reasons, alternatives, RNA support and provenance are preserved. Exploratory predictions can use any hypothesis, but only admitted antigens can enter VaccinePeptide. This adapter expects trusted Isovar output; it does not validate arbitrary caller declarations against reference annotation or discover fusions from BAM files.

from vaxrank import fusion_antigens_from_isovar, predict_epitopes, VaccinePeptide

adapted = fusion_antigens_from_isovar(result, tumor_specificity=attestation)
for antigen in adapted.admitted_antigens:
    epitopes = predict_epitopes(mhc_predictor=predictor, antigen=antigen, genome=genome)
    candidate = VaccinePeptide(
        antigen=antigen, epitopes=epitopes,
        combined_score_expr="target_epitope_score",
        ranking_rules=("target_epitope_score", "manufacturability", "self_epitope_score"),
    )

Use source-agnostic scoring as above, not the small-variant default's n_rna_alt binding. Full reconstruction provenance travels in each antigen's source_metadata; no wild-type protein, self-proteome exclusion or clinical neoantigen status is invented. The existing small-variant run_vaxrank API and the LENS CLI path are unchanged.

Mutant transcript assembly (Isovar)

For each somatic variant, Isovar extracts RNA-seq reads overlapping the mutant locus and assembles them into a mutant protein fragment. This is more accurate than simply applying the DNA variant to the reference transcript because it:

• Phases adjacent germline and somatic variants that fall on the same read, producing the true amino acid sequence
• Captures splicing differences such as intron retention events that may alter the reading frame near the mutation
• Confirms expression — variants with no supporting RNA reads are filtered out

CandidateEpitope scoring

Each mutant protein fragment is sliced into overlapping subsequences of epitope length (typically 8–15 amino acids). These candidate epitopes are scored for predicted MHC binding affinity using the selected predictor. Binding predictions are converted to a score between 0 and 1 via a logistic function parameterised by the EpitopeConfig settings.

Vaccine peptide ranking

Candidate vaccine peptides (longer SLPs, typically 25-mers) are constructed around each mutation. Each candidate is scored by the combined immunogenicity of the epitopes it contains. Candidates are then filtered and ranked by:

1. CandidateEpitope content — total predicted immunogenicity score
2. Reference proteome filtering — peptides matching the human reference proteome are removed to ensure only truly novel sequences are selected
3. Cancer hotspot annotation — variants at known recurrently mutated positions (bundled data from cancerhotspots.org, ~2,700 mutations across cancer types) are flagged
4. Manufacturability — tie-breaking by hydropathy-based synthesis difficulty (C-terminal and 7-mer window GRAVY scores)

Data model

Vaxrank's central data unit is the VaccinePeptide (VP) — one ranked candidate of "this is a vaccine peptide we should consider for this antigen." A VP bundles:

• a source-agnostic VaccineAntigen — the amino-acid sequence, targetable intervals, source provenance, and tumor-specificity attestation;
• for ordinary SNV/indel antigens, a MutantProteinFragment with mutation positions, gene name, source variant, and RNA evidence metrics;
• a list of EpitopePrediction records — per-(k-mer, HLA-allele) MHC binding predictions, sorted into a mutant set (overlapping the mutation, drives ranking) and a wildtype set (cross-reactivity candidates).

The pipeline output is a list of (varcode.Variant, [VaccinePeptide, ...]) tuples — each variant has 1 or more VPs depending on max_vaccine_peptides_per_variant:

ranked_variants_with_vaccine_peptides = [
    (Variant_A, [VP_A1, VP_A2, ...]),    # multiple windows around variant A's mutation
    (Variant_B, [VP_B1]),                 # single SLP for variant B
    ...
]

For each variant, vaxrank can emit multiple alternate constructs:

• --vaccine-peptide-length — controls SLP size and the default RNA context request; --protein-sequence-length or legacy --padding-around-mutation can explicitly override the reconstruction context.
• max_vaccine_peptides_per_variant (config) — controls how many alternate windows per variant make it into the ranked output.
• --peptide-candidates-per-slot / --mrna-candidates-per-slot (CLI) — controls how many VP alternates per variant slot the construct assembler renders into FASTAs.

Reports render one section per variant; within a section, each VP gets its own per-epitope sub-table — column counts can differ per VP (e.g. when pepsickle credibility tagging succeeded for one VP and failed for another, only the successful VP's table shows the processing columns).

Key modules

Shared upstream:

• core_logic.py: Main vaccine peptide selection algorithm
• epitope_logic.py: CandidateEpitope scoring and filtering
• epitope_io.py: LENS / pVACseq / vaxrank-native I/O for epitope predictions
• external_input.py: Synthesize the canonical ranked-vaccine-peptides shape from a LENS / pVACseq report so external-input runs reach the same dispatch as VCF + BAM
• reference_proteome.py: Set-based kmer index for reference proteome filtering (O(1) lookup, built once and cached)
• cancer_hotspots.py: Cancer mutation hotspot annotation
• vaccine_peptide.py: Vaccine peptide scoring and manufacturability
• vaccine_library.py: Shared linker vocabulary + compositional grammar ((BASE)N, GnSm, AnY, An, Gn) with primary-source citations

Vaccine-type-specific (downstream):

• peptide.py: Peptide construct assembly + FASTA / JSON manifest / vendor order-form CSV writers; sub-modes slp / minimal_epitope / multi_epitope
• mrna.py: mRNA construct assembly + three-FASTA / structured manifest / long-format CSV writers. DnaChisel codon optimization, 2A frozen-codon handling, configurable polyA tail (default A120, optional segmented BNT162b2 pattern), per-junction MHC-aware linker swap (issue #247)
• mrna_library.py: mRNA-specific elements (5'/3' UTRs incl. tandem 2× HBB FI; signal peptides HLA-A / HLA-B / tPA / IgK / CD8A / CD28; MITD HLA-A / HLA-B)
• junction_swap.py: Per-junction linker optimizer that minimizes predicted MHC presentation of chimeric k-mers spanning antigen junctions

Reports:

• report.py: Analysis-report generation (ASCII, HTML, PDF, XLSX, CSV, JSON)

Papers & Citations

Vaxrank algorithm:

Rubinsteyn, A., Hodes, I., Kodysh, J. & Hammerbacher, J. Vaxrank: A Computational Tool For Designing Personalized Cancer Vaccines. bioRxiv (2017).

OpenVax pipeline (methods):

Kodysh, J. & Rubinsteyn, A. OpenVax: An Open-Source Computational Pipeline for Cancer Neoantigen Prediction. Methods Mol. Biol. 2120, 147–160 (2020).

PGV001 clinical results:

Saxena, Marron, Kodysh, et al. PGV001, a Multi-Peptide Personalized Neoantigen Vaccine Platform: Phase I Study in Patients with Solid and Hematologic Malignancies in the Adjuvant Setting. Cancer Discovery 15(5), 930–947 (2025).

Saxena, Anker, Kodysh, et al. Atezolizumab plus personalized neoantigen vaccination in urothelial cancer: a phase 1 trial. Nature Cancer 6(6), 988–999 (2025).

BibTeX for the Vaxrank paper:

@article {Rubinsteyn142919,
    author = {Rubinsteyn, Alex and Hodes, Isaac and Kodysh, Julia and Hammerbacher, Jeffrey},
    title = {Vaxrank: A Computational Tool For Designing Personalized Cancer Vaccines},
    year = {2017},
    doi = {10.1101/142919},
    publisher = {Cold Spring Harbor Laboratory},
    URL = {https://www.biorxiv.org/content/early/2017/05/27/142919},
    journal = {bioRxiv}
}

Dependencies

Vaxrank is built on the OpenVax ecosystem:

• pyensembl: Reference genome annotation
• varcode: Variant effect prediction from DNA
• isovar: RNA-based mutant transcript assembly and variant phasing
• mhctools: Unified interface to MHC binding predictors

Other key dependencies:

• msgspec: Configuration serialization (YAML/JSON)
• pandas, numpy: Data processing
• jinja2, pdfkit/weasyprint: Report generation

Legacy flags

For back-compat with older scripts:

• --peptide-mode {slp, minimal_epitope, multi_epitope} is a shorthand for the orthogonal axes:
  • slp ≡ --antigen-content mutation_spanning --peptide-antigens-per-construct 1
  • minimal_epitope ≡ --antigen-content minimal_epitope --peptide-antigens-per-construct 1
  • multi_epitope ≡ --antigen-content mutation_spanning --peptide-antigens-per-construct N

The orthogonal axes are preferred for new designs.

Development

To install Vaxrank for local development:

git clone git@github.com:openvax/vaxrank.git
cd vaxrank
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
pip install -e .
# Examples; adjust release to match your reference
pyensembl install --release 113 --species human
pyensembl install --release 113 --species mouse

Run linting and tests:

./lint.sh && ./test.sh

The test suite is intentionally repository-only and is not included in source distributions. Its integration tests depend on helpers and genomic fixtures that are not package payloads, so run tests from a Git checkout rather than an unpacked PyPI archive.

The first run of the tests may take a while to build the reference proteome kmer index, but subsequent runs will use the cached index.

Reference membership caches are keyed by the installed annotation/sequence content (or actual protein content for custom reference providers), not only species and release. This supports explicit offline PyEnsembl Genome references and prevents a subset from sharing a full-reference cache. Upgrading from species/release-only caches triggers a one-time rebuild; old cache files are not trusted or automatically deleted.

Scripts

• develop.sh: installs the package in editable mode and sets PYTHONPATH to the repo root.
• lint.sh: runs ruff on vaxrank and tests.
• test.sh: runs pytest with coverage.
• deploy.sh: runs lint/tests/build/upload through one Python environment, verifies PyPI artifacts by SHA-256, and safely resumes an interrupted upload from its original artifacts. Deploy is restricted to synchronized main/master.