varcode 5.0.4

Variant annotation in Python

Varcode

Varcode is a library for working with genomic variant data in Python and predicting the impact of those variants on protein sequences.

Installation

You can install varcode using pip:

pip install varcode

You can install required reference genome data through PyEnsembl as follows:

# Downloads and installs the Ensembl releases (75 and 76)
pyensembl install --release 75 76

Example

import varcode

# Load TCGA MAF containing variants from their
variants = varcode.load_maf("tcga-ovarian-cancer-variants.maf")

print(variants)
### <VariantCollection from 'tcga-ovarian-cancer-variants.maf' with 6428 elements>
###  -- Variant(contig=1, start=69538, ref=G, alt=A, genome=GRCh37)
###  -- Variant(contig=1, start=881892, ref=T, alt=G, genome=GRCh37)
###  -- Variant(contig=1, start=3389714, ref=G, alt=A, genome=GRCh37)
###  -- Variant(contig=1, start=3624325, ref=G, alt=T, genome=GRCh37)
###  ...

# you can index into a VariantCollection and get back a Variant object
variant = variants[0]

# groupby_gene_name returns a dictionary whose keys are gene names
# and whose values are themselves VariantCollections
gene_groups = variants.groupby_gene_name()

# get variants which affect the TP53 gene
TP53_variants = gene_groups["TP53"]

# predict protein coding effect of every TP53 variant on
# each transcript of the TP53 gene
TP53_effects = TP53_variants.effects()

print(TP53_effects)
### <EffectCollection with 789 elements>
### -- PrematureStop(variant=chr17 g.7574003G>A, transcript_name=TP53-001, transcript_id=ENST00000269305, effect_description=p.R342*)
### -- ThreePrimeUTR(variant=chr17 g.7574003G>A, transcript_name=TP53-005, transcript_id=ENST00000420246)
### -- PrematureStop(variant=chr17 g.7574003G>A, transcript_name=TP53-002, transcript_id=ENST00000445888, effect_description=p.R342*)
### -- FrameShift(variant=chr17 g.7574030_7574030delG, transcript_name=TP53-001, transcript_id=ENST00000269305, effect_description=p.R333fs)
### ...

premature_stop_effect = TP53_effects[0]

print(str(premature_stop_effect.mutant_protein_sequence))
### 'MEEPQSDPSVEPPLSQETFSDLWKLLPENNVLSPLPSQAMDDLMLSPDDIEQWFTEDPGPDEAPRMPEAAPPVAPAPAAPTPAAPAPAPSWPLSSSVPSQKTYQGSYGFRLGFLHSGTAKSVTCTYSPALNKMFCQLAKTCPVQLWVDSTPPPGTRVRAMAIYKQSQHMTEVVRRCPHHERCSDSDGLAPPQHLIRVEGNLRVEYLDDRNTFRHSVVVPYEPPEVGSDCTTIHYNYMCNSSCMGGMNRRPILTIITLEDSSGNLLGRNSFEVRVCACPGRDRRTEEENLRKKGEPHHELPPGSTKRALPNNTSSSPQPKKKPLDGEYFTLQIRGRERFEMF'

print(premature_stop_effect.aa_mutation_start_offset)
### 341

print(premature_stop_effect.transcript)
### Transcript(id=ENST00000269305, name=TP53-001, gene_name=TP53, biotype=protein_coding, location=17:7571720-7590856)

print(premature_stop_effect.gene.name)
### 'TP53'

If you are looking for a quick start guide, you can check out this iPython book that demonstrates simple use cases of Varcode.

Further reading

Feature guides live in docs/ and on the docs site:

• Effect annotation — how variants become effects, splice outcome representations, pluggable annotators, and structural variants.
• Genotypes and sample-aware queries — per-sample zygosity on multi-sample VCFs.
• VariantCollection transforms — pair_breakends, left_align_indels, and the transform contract.
• CSV round-trip and metadata headers — to_csv / from_csv with #-prefixed provenance headers.
• Error handling — ReferenceMismatchError, GenomeBuildMismatchError, SampleNotFoundError, and raise_on_error=False.
• Germline-aware annotation — niche. Classify somatic variants against the patient's germline-applied transcript when somatic and germline share a codon.

See CHANGELOG.md for the release history.

Effect Types

Every concrete MutationEffect subclass that varcode may emit, grouped by biological context. Each row links to the class definition in varcode/effects/effect_classes.py via a browser text-fragment URL — links survive line-number drift as the source file evolves. Severity ordering across types is set by effect_priority(); the abstract bases — MutationEffect, TranscriptMutationEffect, CodingMutation, NonsilentCodingMutation, SpliceMechanismEffect, StructuralVariantEffect — define the shared interface; MutationEffect and NonsilentCodingMutation have dedicated entries in the API reference, and MultiOutcomeEffect is described in its own section below.

Effects that carry multiple possibilities

Several effects don't have a single deterministic protein-level outcome — splice-signal disruption can resolve as normal splicing, exon skipping, intron retention, or cryptic-site use; an exon-edge variant might be a routine coding effect or a splice disruption; a structural variant might affect multiple transcripts or have multiple plausible breakpoint resolutions; two or more cis variants on one transcript can compose into a joint mutant protein; and when phase is unknown between somatic and germline variants sharing a window, the somatic effect depends on which haplotype it landed on. Varcode wraps these as MultiOutcomeEffect instances. Every multi-outcome effect exposes the same surface:

• .candidates → tuple[EffectCandidate, ...]
• Each EffectCandidate wraps an inner effect (.effect), a producer tag (.source, e.g. "varcode", "rna_evidence"), and free-form .evidence.
• .effects → tuple[MutationEffect, ...] — convenience that unwraps .candidates to inner effects when provenance isn't needed.
• .most_likely_candidate / .most_likely_effect — producer-ordered top pick.
• .highest_priority_candidate / .highest_priority_effect — most severe by effect_priority().

The MultiOutcomeEffect containers appear in the sub-tables below where they're emitted: SpliceOutcomeSet, ExonicSpliceSite, the StructuralVariantEffect sub-hierarchy, HaplotypeEffect, and PhaseCandidateSet.

Coding region — in-frame changes

Effect type	Description
Substitution	Coding mutation which causes simple substitution of one amino acid for another.
Insertion	Coding mutation which causes insertion of amino acid(s).
Deletion	Coding mutation which causes deletion of amino acid(s).
ComplexSubstitution	Insertion and deletion of multiple amino acids.
Silent	Mutation in coding sequence which does not change the amino acid sequence of the translated protein.
AlternateStartCodon	Replace annotated start codon with alternative start codon (e.g. "ATG>CAG"); a Silent subclass since the initiator tRNA still loads Met.

Coding region — frame-disrupting / truncating

Effect type	Description
FrameShift	Out-of-frame insertion or deletion of nucleotides, causes novel protein sequence and often premature stop codon.
FrameShiftTruncation	A frameshift which leads immediately to a stop codon (no novel amino acids created).
PrematureStop	Insertion of stop codon, truncates protein.
StartLoss	Mutation causes loss of start codon, likely result is that an alternate start codon will be used down-stream (possibly in a different frame).
StopLoss	Loss of stop codon, causes extension of protein by translation of nucleotides from 3' UTR.

Splice-site disruption — where the signal was hit

DNA-level locations: these effects say a variant landed on or near a splice signal, but say nothing about how the spliceosome responds (see the next table for that).

Effect type	Description
SpliceDonor	Mutation in the first two nucleotides of an intron, likely to affect splicing.
SpliceAcceptor	Mutation in the last two nucleotides of an intron, likely to affect splicing.
IntronicSpliceSite	Mutation near the beginning or end of an intron but less likely to affect splicing than donor/acceptor mutations.
ExonicSpliceSite	Mutation at the beginning or end of an exon, may affect splicing; itself a MultiOutcomeEffect wrapping the alternate exonic coding effect alongside the splice candidates.

Splice mechanism — what the spliceosome does in response

The protein-level consequence of a splice-signal hit is not deterministic from DNA alone, so varcode emits these as candidates inside a SpliceOutcomeSet (a MultiOutcomeEffect). Each mechanism carries the originating disruption on its .splice_signal attribute, so you can always recover where the hit was off any mechanism.

Effect type	Description
NormalSplicing	Splice signal hit but splicing proceeds normally; protein consequence (if any) is whatever the underlying nucleotide change would produce.
ExonSkipping	Affected exon excluded from the mature transcript; in-frame skip deletes amino acids, out-of-frame skip propagates a frameshift.
IntronRetention	Intron stays in the mature transcript; translation usually hits a premature stop inside the retained intron.
CrypticDonor	Disrupted canonical donor replaced by a nearby cryptic GT donor; exon extended or truncated.
CrypticAcceptor	Disrupted canonical acceptor replaced by a nearby cryptic AG acceptor; exon extended or truncated.

Non-coding regions and unclassifiable contexts

Effect type	Description
FivePrimeUTR	Variant affects 5' untranslated region before start codon.
ThreePrimeUTR	Variant affects 3' untranslated region after stop codon of mRNA.
Intronic	Variant occurs between exons and is unlikely to affect splicing.
NoncodingTranscript	Transcript doesn't code for a protein.
IncompleteTranscript	Can't determine effect since transcript annotation is incomplete (often missing either the start or stop codon).
Intergenic	Occurs outside of any annotated gene.
Intragenic	Within the annotated boundaries of a gene but not in a region that's transcribed into pre-mRNA.
Failure	Placeholder effect emitted when annotation failed but a non-empty effect list is required (raise_on_error=False).

Exon-level and structural-variant effects

ExonLoss is a plain Exonic effect. The structural-variant effects below (LargeDeletion through TranslocationToIntergenic) are MultiOutcomeEffects — their .candidates may include cryptic-exon outcomes, RNA-evidence-ranked alternatives, and so on. CrypticExonCandidate typically appears as a candidate inside those SV effects rather than standalone.

Effect type	Description
ExonLoss	Deletion of an entire exon, significantly disrupts protein.
LargeDeletion	Structural deletion (<DEL> / <CN0>) removing one or more exons or an entire gene.
LargeDuplication	Tandem duplication (<DUP>) overlapping exons; may yield copy-number increase or a fused reading frame.
Inversion	Inversion (<INV>) flipping a stretch of a transcript; consequence depends on whether breakpoints fall in exons or introns.
GeneFusion	Breakend (<BND>) whose mate lies in another protein-coding gene — the canonical fusion shape.
TranslocationToIntergenic	Breakend whose mate lies in intergenic space; consequence depends on cryptic splice / ORF signals downstream.
CrypticExonCandidate	An SV brings novel sequence into range of a transcript and motif scoring flags a plausible new splice acceptor / donor pair; attached as additional candidates on SV effects.

Multi-variant / phase-dependent effects

Both are MultiOutcomeEffects, emitted alongside per-variant effects (additive, not a replacement) when a phase resolver groups cis variants together or when phase between somatic and germline variants is unknown.

Effect type	Description
HaplotypeEffect	Joint effect of two or more cis variants on the same transcript; the combined mutant cDNA is built and translated as one unit.
PhaseCandidateSet	Possibility set across phase hypotheses when a somatic variant and one or more germline variants share a window on a transcript and phase between them is unknown.

Coordinate System

Varcode currently uses a "base counted, one start" genomic coordinate system, to match the Ensembl annotation database. We are planning to switch over to "space counted, zero start" (interbase) coordinates, since that system allows for more uniform logic (no special cases for insertions). To learn more about genomic coordinate systems, read this blog post.