The source's id for one credible set from one study (Open Targets' studyLocusId, a 32-hex hash). Part of this row's identity. Opaque: it is a join key back to the release, and it changes when the source re-runs fine-mapping.
gene_id
str
required
Ensembl gene accession of the candidate gene (ENSG…), as the source publishes it. Part of the identity. No symbol: the source's L2G table carries none, and a symbol looked up elsewhere would be another source's claim on this row.
variant_key
str
required
The module's key for the variant the pass asked about, a member of this credible set (its lead or not): part of the identity, so two module variants in one set each get the set's rows. Joins weights.parquet.variant_key directly; the member's own weight in the set is its gwas_effects.csv posterior.
lead_variant_id
str | None
optional
The source's own spelling of the credible set's lead variant (19_44908684_T_C, or an OTVAR_… hash for a long allele), kept for tracing back to the release. Names another variant when variant_key is a member and not the lead. Never parsed: the alleles come from the source's columns, not from this id.
study_accession
str | None
optional
The study the credible set belongs to (GCST… or a FinnGen id), the same column gwas_effects.csv carries, so the two tables join on (variant_key, study_accession).
l2g_score
float | None
optional
Open Targets' locus-to-gene score for this pair, 0 to 1, verbatim. A model output, retrained each release, and comparable only within one dataset. The method is in the name because a score from another gene-prioritisation method is on another scale and would be a column of its own. Not an effector call: what score makes a gene the effector is the module's rubric, never this table's.
dataset
str
required
Which release this row came from, e.g. 'opentargets_26.09'. A FACT: L2G is retrained per release.
source
str | None
optional
The licensed data source: opentargets|manual (open). Joins sources.csv.source.
fetched_at
str | None
optional
ISO-8601 UTC timestamp when a pass wrote this row; the release is dataset.
Generated from the row model at build time — reference.authoring_reference(), the same answer describe_table gives an authoring tool. Nothing on this page is hand-kept.