just_dna_format.ld_pairs¶
just_dna_format.ld_pairs ¶
The computed LD pair behind a study measured on a proxy (0.8, RM331).
ld_pairs.csv is a derived-fact sidecar filled by just-dna-enricher ld-pairs. An author writes only
StudyRow.proxy_rsid, the variant a study measured in place of the row's lead: r² has no meaning without
the population it was computed in, so r² and its population are a pair, and the pair is computed from a
reference panel rather than typed (the maintainer's cut, 2026-11-04). One row per
(proxy_rsid, variant_key, ld_population).
The population is the reference's name for its panel, not the study's code. 1000GENOMES:phase_3:EUR
is what Ensembl computed r² in; the study's ancestry is EUR. Keeping the panel's own spelling says
which reference the number is from without a second column, and leaves a different panel (another
release's, another reference's) a different key rather than the same word meaning two things.
Withheld is a status, never a zero. Ensembl answers 200 [] both for a pair on two chromosomes and
for a pair whose r² falls under an internal floor, and lowering its r2 parameter brings no row back
(docs/probes/LANES_0_8.md §2.4). That answer is not_reported, with r2 empty. A study naming several
superpopulations, or none, gets its row too, with the reason and no population: the lookup never picks a
panel for it.
LdPairRow ¶
Bases: BaseModel
One (proxy, lead, panel): r² and D′ between them, or the reason they are withheld.