Propositions (Steps 2--6)¶

Overview¶

Steps 2--6 of gkm_scv_statement_proc build the qualifier tables and assemble the primary and target propositions for each SCV. Qualifiers capture gene context, mode of inheritance, and penetrance; propositions combine the variant, condition, and qualifiers into the structured assertion that forms the core of each VA-Spec Statement.

Steps 2--4: Qualifier Tables¶

Step 2: Build `temp_gene_context_qualifiers`¶

Extracts gene context from the upstream single_gene_variation table: one representative gene per variant, chosen by the single-gene method in clinvar-ingest-bq-tools. The gene's symbol and hgnc_id come from that same table, so this step joins no gene table of its own. For each SCV whose variant has a row, it builds a gene concept with:

conceptType: "gene"
primaryCoding: NCBI Gene identifier using both identifiers.org and NCBI Gene URLs
HGNC mappings: When an HGNC identifier is available, included as an additional coding
submittedGeneSymbols extension: Gene symbols from clinical_assertion_variation, preserving the submitter's original gene annotations. Trimmed of the space a ,-separated list leaves after the split, distinct, in submitted order: variation record, then gene list, then position in the comma-separated symbol list. That key is unique, so the order is total and an incremental build matches a full one.
geneContextSource extension: Where the gene came from. The submitter's symbols never choose the gene; they only set this value. They are compared with the gene's symbol after trimming spaces and ignoring case:
- submittedSymbolMatches — the submitter's symbols include the gene's symbol
- submittedSymbolDiffers — the submitter gave symbols, none equal to it (mostly previous symbols of the same gene)
- inferredFromVariant — the submitter gave no symbols

When no single-gene match exists for a variation but the submitter gave symbols, the qualifier is a placeholder. Its name is the submitter's symbol: several are joined with , in submitted order. It has no primaryCoding or mappings, keeps submittedGeneSymbols, and has geneContextSource = submittedNotNormalized. Releases before 1.1.0 name the placeholder submitted genes were not normalized.

The schema accepts only these four geneContextSource values. ClinvarGeneContextQualifier defines them, and the bundle schema applies it to every proposition type, including the VA-Spec ones.

The extension values are stored as typed fields (value_string for the symbol array, value_enum for the source) and collapsed to value at publish.

Output: temp_gene_context_qualifiers -- one row per SCV+gene combination. Internal

Step 3: Build `temp_moi_qualifiers`¶

Extracts ModeOfInheritance from assertion attributes in clinical_assertion. When a matching HPO term is available, the qualifier includes a primaryCoding with the HPO term identifier and label. All qualifiers include a submittedModeOfInheritance extension preserving the original submitted value.

Output: temp_moi_qualifiers -- one row per SCV with mode of inheritance. Internal

Step 4: Build `temp_penetrance_qualifiers`¶

Derives penetrance qualifiers for specific classification types:

Classification Category	Penetrance Value
Pathogenic-low penetrance (`p-lp`), Likely pathogenic-low penetrance (`lp-lp`)	`"low"`
Established risk allele (`era`), Likely risk allele (`lra`), Uncertain risk allele (`ura`)	`"risk"`

Each penetrance qualifier includes a submittedClassification extension preserving the original classification label that triggered the penetrance derivation.

Output: temp_penetrance_qualifiers -- one row per qualifying SCV. Internal

Step 5: Primary Proposition¶

Assembles the SCV proposition by joining temp_gkm_scv with all qualifier tables and condition sets from gkm_scv_condition_sets. The resulting proposition contains:

Field	Description
`type`	Proposition type from Step 1 mapping (e.g., `VariantPathogenicityProposition`)
`subject`	Reference to the categorical variant via `clinvar:{variation_id}`
`predicate`	Predicate from Step 1 mapping (e.g., `isCausalFor`, `isOncogenicFor`)
`object_single`	Single condition from the condition pipeline
`object_compound`	ConditionSet for SCVs with multiple conditions
`geneContextQualifier`	Gene concept from Step 2
`modeOfInheritanceQualifier`	Mode of inheritance from Step 3
`penetranceQualifier`	Penetrance from Step 4

Output: temp_gkm_scv_proposition -- one row per SCV with fully assembled proposition. Internal

Step 6: Target Proposition (Somatic)¶

Builds the evidence line target proposition for somatic clinical impact assertions. This proposition uses the evidence_line_target_proposition type and predicate derived in Step 1 and adds somatic-specific fields:

Field	Description
`type`	Target proposition type (e.g., `VariantPrognosticProposition`, `VariantTherapeuticResponseProposition`)
`subject`	JSON pointer `4/proposition/subject` referencing the parent proposition's variant
`predicate`	Target predicate (e.g., `associatedWithBetterOutcomeFor`, `predictsSensitivityTo`)
`object`	For therapeutic assertions, a reference to the therapy: `#/therapy/clinvar.therapy:{sha256}` for a single drug, or `#/therapyGroup/clinvar.therapygroup:{sha256}` for a multi-drug combination. Therapies are extracted (`temp_gkm_scv_therapy`), content-addressed, and deduplicated into the `gkm_dict_therapy` / `gkm_dict_therapygroup` dictionaries; a TherapyGroup's `concepts` reference its member therapies via `#/therapy/`
`conditionQualifier`	Condition moved to qualifier position for therapeutic assertions (since `object` becomes the therapy), as a `#/condition/` or `#/conditionSet/` reference
`geneContextQualifier`	Gene concept from Step 2
`modeOfInheritanceQualifier`	Mode of inheritance from Step 3

The JSON pointer 4/proposition/subject is used instead of duplicating the variant reference, linking the target proposition back to the same variant defined in the parent (Step 5) proposition.

Output: temp_gkm_scv_target_proposition -- one row per somatic SCV with target proposition. Internal

Dependencies¶

Source Tables: single_gene_variation (with its symbol and hgnc_id columns), clinical_assertion, clinical_assertion_variation
Lookup Tables: hpo_terms
Upstream Steps: Step 1 (temp_gkm_scv), condition pipeline (gkm_scv_condition_sets)