Curating noncoding variant impact

Use the existing fields in the schema. Keep classifications small and optional; preserve evidence-specific detail in descriptions. A noncoding location alone does not establish a regulatory mechanism. Splicing, RNA stability, translation, and noncoding-RNA defects may require different causal steps. An SV can also overlap coding genes while acting primarily through regulation.

Keep the claims separate

Claim	Where to record it
Physical alteration	Variant.variant_type; initiating node's genetic_context.variant_type
Sequence features overlapped	genomic_contexts on either of those objects
Named affected region and its position	affected_regions on Variant, GeneticContext, or a regional Genetic record
Gene whose expression is affected or proposed to be affected	Variant.regulatory_target_gene; gene annotation on the downstream expression node
Element and direct molecular effect	functional_effects[].regulatory_element_type, regulatory_mechanism, and description
Expression pattern	regulatory_category on Variant, FunctionalEffect, or Pathophysiology, when supported
Quantitative process change	Biological-process descriptor modifier: INCREASED or DECREASED
Downstream consequences	Separate, evidence-backed pathophysiology nodes and causal edges

variant_type and genomic_contexts use human-readable static enum values, such as deletion, single nucleotide variant, intron, or "5' UTR". The Sequence Ontology mappings live in the schema; do not add ID/label pairs to these fields. Gene and process annotations still use their normal descriptors.

Contexts can overlap and depend on the transcript. Identify the relevant genes/transcripts in description; an intronic host gene may differ from the regulatory target. Prefer the appropriate UTR context when the overlap is known to be untranslated sequence of a protein-coding transcript, including a wholly untranslated exon such as APC exon 1B. Use noncoding exon for an entire exon without codons when no more specific UTR context applies. Omit unknown contexts. Only put contexts shared by the represented alleles on a shared initiating node. These classifications neither identify an enhancer nor encode genomic coordinates.

regulatory_target_gene does not assert sequence overlap or prove causality. When a variant overlaps the gene and affects its regulation, both gene and regulatory_target_gene may be appropriate. For an intact target, avoid putting that gene on the physical lesion node as though its sequence were disrupted.

Keep legacy type and allele_type valid alongside the controlled fields; do not migrate unrelated entries or require both representations. Rendering and exports prefer variant_type while retaining distinct legacy detail. For a complex alteration outside the enum, retain free text rather than forcing a class.

Describe affected regions qualitatively

Use optional affected_regions when a named enhancer, boundary, or chromosomal interval conveys more than a gene list. A GenomicRegion requires only name; add description for its scope, regulatory_element_type when known, and chromosomal_region for a reported cytoband or band range such as 7q36 or 16p12.2-p11.2. Do not require an ontology identifier or one set of coordinates for a disease whose alleles differ. Keep legacy descriptions valid; annotate regions as entries are curated rather than migrating unrelated records.

The gene relationships locate the named region in the linear reference genome. They use ordinary GeneDescriptor objects, with verified HGNC IDs when available:

Slot	Meaning
between_genes	Exactly two distinct, unordered gene landmarks on opposite sides of the region, with no overlap of either anchor; neither the nearest genes nor exact interval endpoints are implied
within_gene	One gene whose genomic span contains the named region, for example ZRS within LMBR1
overlaps_genes	Genes whose genomic spans overlap the named region; the list need not be exhaustive
adjacent_to_genes	Genes that share a sequence boundary with the region without overlapping it; do not use for vaguely nearby genes

These slots do not assert a regulatory target, a causal gene, a chromatin contact, or adjacency created by a rearrangement. Omit a relationship the source does not establish, and do not assign contradictory spatial relations to the same gene (such as both within and strictly adjacent). There are no left/right slots: genomic coordinate direction and transcriptional direction must not be conflated. A phrase such as "upstream of SHH" can remain in the description when the evidence supports that detail but not one of the available spatial relations.

Choose the subject of each annotation carefully. An EPHA4-PAX3 boundary lies between gene landmarks; an entire deletion that also removes EPHA4 does not. Likewise, ZRS lies within LMBR1, but a ZRS-encompassing duplication may extend beyond LMBR1. The affected_regions list can identify the relevant element within a larger alteration without claiming to exhaust that alteration.

affected_regions:
- name: EPHA4-PAX3 regulatory boundary
  regulatory_element_type: TAD_BOUNDARY
  between_genes:
  - preferred_term: EPHA4
    term:
      id: hgnc:3388
      label: EPHA4
  - preferred_term: PAX3
    term:
      id: hgnc:8617
      label: PAX3
  description: >-
    The boundary is deleted together with EPHA4 coding sequence;
    PAX3 coding sequence remains intact. The landmarks locate the
    boundary, not the full deletion interval.

Support the region annotations in the enclosing variant, mechanism node, or genetic record's evidence; GenomicRegion has no separate evidence slot. For a regional Genetic record, omit gene_term if no gene-level causal association is being asserted. A host gene or flanking landmark belongs in the region object, not in a substitute causal-gene binding. Retain a Variant.gene annotation when the variant actually overlaps that gene, and keep regulatory_target_gene separate. In exports, positional gene landmarks must remain region metadata rather than becoming causal-gene edges.

The worked examples are Preaxial Digit Brachydactyly-Webbed Fingers, ZRS-Related Limb Malformation, and Chromosome 16p12.2-p11.2 Deletion Syndrome.

Classify expression effects conservatively

Category	Meaning in the current schema
LOE	Reduced or absent expression across all normally expressing cell types
mLOE	Reduced or absent expression in a subset of cell types or developmental windows
GOE	Ectopic spatial or temporal expression

Increased expression in normally expressing cells is not sufficient for GOE. Use gene-expression modifier: INCREASED or DECREASED for measured abundance changes. A reduction in one assayed tissue may leave LOE versus mLOE unresolved; leave the category unset and describe that limit. The enum's coding LOF, GOF, and DN values do not substitute for expression effects. Use genetic_context.functional_impact_category for a supported variant functional consequence, independently of the expression-pattern classification.

Classification and confidence are independent. For example, regulatory_category: GOE with mechanism_confidence: HYPOTHETICAL on an expression node represents proposed ectopic expression; omitted confidence defaults to established. Variant has no mechanism-confidence slot: qualify uncertain effects in its descriptions and evidence rather than copying a hypothetical node's category as an established fact.

Build the causal account from the evidence

Separate the physical variant, altered regulatory interaction, expression change, and downstream disease mechanism into atomic nodes where supported. Classify the initiating node through genetic_context; separate deletion and inversion triggers when their classes or overlaps differ. Distinguish loss of silencer DNA from loss of silencer contact or control when the element remains present.

Record assayed tissue, cell type, developmental stage, and experimental system. Expression measurements do not by themselves establish altered transcription; RNA stability may also explain them. Computational contact predictions and cross-species assays need their own evidence grading, distinct from patient measurements. Normal expression in blood does not automatically refute a tissue-specific mechanism. Preserve possible contributions from other genes in a multigene SV rather than attributing every phenotype to the regulatory target.

For individual alleles, retain the reported genome build and coordinate precision in the description: array probe bounds are not nucleotide-resolved breakpoints. Use dismech-references for source verification and exact snippets, and dismech-terms for ontology bindings. FunctionalEffect has no evidence slot; support its claims on the variant and the corresponding mechanism nodes. Follow the normal curation history and validation workflow in CLAUDE.md.