This glossary covers both the concepts returned by gifter and the technical language used to curate and compile its reference database. Resource-specific terms describe how gifter interprets them; they are not replacements for the resources’ own documentation.
The short form of the evidence hierarchy is:
Frame --resolves by metadata--> GIFT
metabolic GIFT
-> route -> reaction -> enzyme system -> component -> marker
structural, regulatory or defense GIFT
-> architecture, circuit or mechanism
-> function -> system -> component -> markerA frame begins the analytical browsing path, but it is not a biological parent of a GIFT and stores no fixed list of GIFT identifiers.
GIFT and analysis concepts
Anchor
A curated input or output molecule defining the boundary of a metabolic GIFT. Only declared anchors connect GIFTs. An internal reaction participant is not an anchor merely because it occurs in the chemistry.
Architecture
One complete curated implementation of a structural GIFT. Architectures are alternatives under OR logic; every required function within one architecture is combined under AND logic.
Assessability
Whether a genome is complete enough for its lack of support for a GIFT to be informative. Assessability may change a denominator or make an absence indeterminate. It can never turn an unsupported GIFT into a supported one.
Bounded frame
A frame whose curation deliberately intends to enumerate a biologically closed set. Only a bounded frame licenses a coverage fraction over all its members. An open and growing catalogue is not bounded.
Compartment
The extracellular or cytoplasmic qualifier
on a declared anchor. It makes transport and possible cross-genome
compatibility explicit without introducing compartment-aware
stoichiometry or inferring protein localization.
Complete
The Boolean state reached when at least one entire curated implementation of a GIFT is supported. Completeness means encoded capability under the model, not expression, activity, flux or phenotype.
Component
One required protein role within an enzyme or machinery system. Every required component of a system is combined under AND logic; any accepted marker may support a component under OR logic.
Facet
A curated classification used to describe or select GIFTs, anchors or
related entities. A facet does not decide the completeness model;
gift_type does.
Frame (reference frame)
The comparison set and denominator within which a quantitative trait
is read. A named frame stores metadata filters and interpretation
limits, then resolves its current GIFT members against a database
release. It never stores a list of gift_id values.
Function
In a structural, regulatory or defense model, one required biological role within an architecture, circuit or mechanism. It is the machinery analogue of a metabolic reaction and is implemented by any one complete system.
GIFT
A genome-inferred functional trait: a biologically meaningful capability whose genomic support is evaluated through an explicit, curated and traceable completeness model.
GIFT type
One of metabolic, structural,
regulatory or defense. It selects the
completeness model and determines which source tables may attach to a
GIFT. It is part of the core ontology, not a facet.
Internal intermediate
A reaction participant inside a metabolic route that is not a declared anchor. Shared internal intermediates do not connect GIFTs.
Marker
A namespaced genomic observation used as evidence for a component.
Marker identity is the pair namespace + accession; a marker
is never itself a GIFT, reaction or function.
Mode
The direction or role of a metabolic GIFT, such as
anabolic, catabolic, transport or
interconversion. Non-metabolic GIFTs do not have a
mode.
Reaction
A biochemical transformation, identified by its stable
reaction_id and by a Rhea master wherever Rhea covers the
chemistry. Its contextual direction is stored on route membership rather
than by changing reaction identity.
Route
A curated minimal set of required reactions connecting a metabolic GIFT’s declared input and output anchors. Routes are alternatives under OR logic; all required reactions within a route are combined under AND logic.
Supported
Backed by the observed markers at the relevant level of the completeness hierarchy. Support propagates only through the declared OR/AND rules. It is not evidence of expression or activity.
Marker and profile terminology
Accession
A stable identifier assigned by an external resource, such as
K00764, PF01752 or NF040708.3. In
gifter it is interpreted together with its namespace. Similar-looking
strings from different namespaces are not interchangeable.
Bit score
A log-odds score describing how strongly a sequence matches a profile HMM. Scores are meaningful only with the relevant model, thresholds and coverage; gifter does not treat a high score alone as proof that a marker is suitable for a biological claim.
Domain
A conserved region of a protein that may occur in proteins with
different overall functions. A domain-level match often supports
structural homology but not the specific activity of a complete protein.
In NCBIfam, the domain family type does not carry the
one-function assertion required for automatic admission by grade.
Equivalog
An NCBIfam family type asserting that the proteins recognized by the
profile share one specific function. gifter admits
equivalog as an eligible grade, but still checks the
biological annotation, reaction direction, taxonomic scope and
specificity before accepting the profile as marker evidence.
Equivalog domain
An NCBIfam profile covering a function-specific domain rather than necessarily the full protein. It carries NCBIfam’s one-function assertion and is an eligible grade in gifter, but we must still verify that the domain is sufficient evidence for the component being claimed.
Gathering threshold
The model-specific score cutoff curated for accepting members of a profile HMM family. It should be used with the profile release it belongs to; it is not a universal score cutoff shared across models.
Homolog
A sequence related to another sequence by common ancestry. Homology alone does not establish that two proteins perform the same function.
Ortholog
Homologous genes related through speciation rather than duplication. Orthology can support functional inference, but an orthology group may still be broader than a reaction, substrate or machinery role curated by gifter.
PfamAutoEq
An automated NCBIfam family type derived from a Pfam model. gifter
does not treat it as an admitted equivalog grade; the label does not
supply the curated one-function guarantee required for
NCBIFAM marker admission.
Profile HMM
A profile hidden Markov model representing position-specific patterns learned from a multiple sequence alignment. Searches produce scores and alignments against which model-specific thresholds and coverage requirements are applied.
Sequence coverage
The fraction of the profile or target sequence included in an alignment. It helps distinguish a complete family match from a short local match but does not by itself establish functional specificity.
External identifiers and resources
CAZy and dbCAN
CAZy classifies carbohydrate-active enzymes into families and, where available, subfamilies. dbCAN provides sequence models and annotation tools for those classes. A broad CAZy family can contain several characterized activities, so family membership is not automatically evidence for one substrate-specific GIFT.
ChEBI identifier
An identifier for a chemical entity. gifter uses ChEBI identifiers for declared anchors where possible; it does not use them to turn every internal metabolite into an anchor.
EC number
An Enzyme Commission classification of catalysed activity. An EC number may be incomplete or may group chemistry more broadly than one curated Rhea reaction, so it is evidence and cross-reference material rather than a substitute for reaction curation.
KEGG Orthology (KO)
A KEGG functional orthology group identified by a K
accession. KOs are a major marker namespace in gifter, but each KO is
checked against the specificity of the component and reaction it is
proposed to support.
Namespace
The identifier system in which an accession has meaning, such as
KO, EC, PFAM,
TIGRFAM, NCBIFAM or CAZY. Marker
identity is always the namespace and accession together.
NCBIfam
NCBI’s collection of protein family profile HMMs used by tools
including PGAP and exposed through other annotation systems. gifter pins
a named release, requires versioned accessions, and admits only the
equivalog and equivalog_domain grades after
biological review.
Pfam
A database of protein families and domains represented by profile HMMs. A Pfam match can be strong evidence of family or domain membership without necessarily identifying the precise activity required by a GIFT.
Curation and database-generation terms
Candidate
A capability, route, architecture or marker under investigation. A candidate is not part of the ontology and cannot produce a call until it has passed curation, validation and release.
Confidence
A curated description of the strength or ambiguity of accepted evidence. Confidence summarizes evidence quality; it does not replace the Boolean completeness result or promote an unsupported GIFT.
Database source
The human-reviewable TSV files from which the runtime SQLite database is compiled. They are the biological source of truth; SQLite is a generated artifact.
Deferral
A decision that a candidate is worth revisiting but is presently blocked. A deferral records the blocker and the evidence or resource change that would retrigger investigation.
Diagnostic proxy
A provisional marker set used to measure prevalence, disagreement or likely failure modes during curation. A proxy is not a GIFT definition and is never silently promoted into one.
Evidence type
The kind of relationship supporting a component-to-marker assertion, such as orthology or sequence-family evidence. It records what the marker represents; it is distinct from evidence confidence.
Marker specificity
How narrowly a marker identifies the component or chemistry being claimed. The specificity of a GIFT may never exceed the specificity of the evidence that supports it.
Marker-specificity screen
A reproducible comparison used to identify over-broad accessions, coverage gaps, conflicts between resources or profiles worth manual review. A screen creates evidence for curation; it does not automatically admit its candidates.
Provenance
The recorded origin of a biological fact or curation decision: resource, release, retrieval date, literature or gifter-specific interpretation. Imported facts are kept distinct from gifter’s choices of boundaries, implementations and claim scope.
Refusal
A recorded decision not to curate a candidate because the evidence cannot support the proposed claim or the claim violates the ontology. A refusal is a durable curation result rather than an undocumented gap.
Retrigger
A concrete change that justifies reopening a deferred or refused assessment, such as a new specific marker, characterized chemistry or a resource release that resolves the recorded blocker. Repeating the same request is not a retrigger.
SQLite artifact
The compiled, read-only runtime representation of the curated source tables. It is regenerated after source changes and is never edited manually.
For the end-to-end process that connects these terms, see How a GIFT is built.