Skip to content

Schema diagram

Generated 2026-09-22 from model/schema/ber_feature_model.yaml with just diagram, using LinkML's erdiagramgen and the model-specific cardinality corrections in scripts/schema_diagram.py. Regenerate after any schema change; this file is not auto-updated. just check compares the generated Mermaid block with this committed block in CI.

erDiagram
AlignmentTarget {
    string gap
    integer target_end
    string target_id
    integer target_start
    StrandEnum target_strand
}
Attribute {
    uriorcurie attribute_cv_id
    string key
    float numeric_value
    string value
}
Contig {
    string contig_id
    string generated_by
    integer length_bp
    float lineage_confidence
    uriList source_files
    stringList taxonomic_lineage
    TopologyEnum topology
    integer translation_table
}
ContigCollection {
    string name
    string collection_id
    ContigCollectionTypeEnum collection_type
    string generated_by
    uriList source_files
    stringList taxonomic_lineage
}
Dataset {

}
FeatureLocation {
    boolean crosses_origin
    LocationOperatorEnum location_operator
}
LocationPart {
    integer end
    EndpointStatusEnum end_status
    integer start
    EndpointStatusEnum start_status
    StrandEnum strand
}
SequenceFeature {
    string name
    CoordinateSystemEnum coordinate_system
    stringList dbxref
    integer end
    string feature_id
    string generated_by
    boolean is_representative
    boolean is_selected
    MethodEnum method
    stringList note
    stringList ontology_term
    integer phase
    string product
    string product_source
    float score
    ScoreTypeEnum score_type
    string source
    uriList source_files
    stringList stable_identifiers
    integer start
    StrandEnum strand
    string translated_sequence
    string type
}

Contig }o--o{ ContigCollection : "member_of"
Dataset ||--o{ Contig : "contigs"
Dataset ||--o{ ContigCollection : "contig_collections"
Dataset ||--o{ SequenceFeature : "features"
FeatureLocation ||--|{ LocationPart : "parts"
Contig ||--o{ LocationPart : "seqid"
SequenceFeature ||--o| AlignmentTarget : "target"
SequenceFeature ||--o| FeatureLocation : "location"
SequenceFeature ||--o{ Attribute : "attributes"
SequenceFeature }o--o{ SequenceFeature : "parent"
Contig ||--o{ SequenceFeature : "seqid"
SequenceFeature ||--o{ SequenceFeature : "seqid (protein coordinates)"

Dataset is the tree root, added 2026-09-18 so a whole harmonized data file validates as one instance of one class instead of a bag of loose Contig and SequenceFeature fragments. Its own box is empty because it has no scalar slots, only the two relationships shown.

erdiagramgen lists every scalar and enum slot inside each entity box and draws an arrow only for slots whose range is another class (Contig, SequenceFeature, Attribute). model/schema/ber_feature_model.yaml itself is the source of truth if this diagram and that file ever disagree.

Each SequenceFeature references exactly one Contig, while a Contig may have zero or many Features. Parent links are optional and many-to-many: a child can have multiple parents, and a parent can have multiple children. Each optional FeatureLocation contains one or more LocationParts; every part references one Contig, which may have many parts. The wrapper preserves those inverse cardinalities when regenerating; it fails for review if the relevant schema constraints or upstream generator output change. It does not modify the schema.