HPC.LMPP: Studies

SCREATE tier definitions supplied by the atlas curator. This methods note explains the prepared classifications; it does not recompute or change any report scores or tiers.

Cell–trait tiers

Cell–trait links

Identify cell types whose regulatory activity is associated with a trait’s genetic signal.

Required for Specific or Relevant

The ICE-CREAM score derived from the mean metacell Z score must be ≥ 5.39, with BOTH nominal and adjusted ICE-CREAM P values ≤ 0.05.

Specific (Tier 1)

Also meets all cell-set enrichment analysis (CSEA) criteria in at least one tested comparison: nominal P ≤ 0.05 AND adjusted P ≤ 0.05 AND normalized enrichment score (NES) ≥ 1, all in that comparison. This indicates that trait-associated activity is concentrated in this cell type relative to the tested background.

Relevant (Tier 2)

Meets the ICE-CREAM requirements but not the additional CSEA criteria. The cell type may share trait-associated activity with other cell types.

Interpretation and report values

CSEA is cell-set enrichment analysis, analogous to gene-set enrichment analysis but using sets of cells along a ranked list of ICE-CREAM scores. The eligibility score derived from the mean metacell Z score is not the displayed mean IC score. Do not infer a tier from the displayed mean, its descriptive Significance column, or a single CSEA summary value. Reports preserve prepared tiers; Other means an available association outside Specific/Relevant, and unavailable is not zero or proof of biological irrelevance.

Meta traits and individual traits

Use the prepared tier for the exact meta trait or individual trait. A meta trait’s Best child CSEA FDR is the lowest available child FDR, not a combined or meta-trait significance test. Its displayed mean child NES is descriptive, not a new CSEA test. These summaries must not be used to reconstruct or override the meta-trait tier.

Gene–trait priorities (variant–tCRE–gene evidence)

Variant–tCRE–gene links

Candidate variants connect to genes through overlapping promoters or distal transcribed cis-regulatory elements (tCREs) linked to gene promoters by the Activity-by-Contact (ABC) model.

Required genetic support

The variant belongs to a purity-passing SuSiE credible set OR qualifies for Borzoi rescue. Borzoi rescue requires BOTH a variant-causality score ≥ 0.3 AND linkage disequilibrium with the lead variant of r² ≥ 0.2.

Required cellular support

The overlapping promoter has positive tCRE activity OR the distal element has a passing cell-specific ABC link to the gene, in a Specific or Relevant cell subtype (L1S) or fine-grained state (L2). BOTH genetic and cellular support are required for High or Medium priority.

High priority

Among records meeting both minimum requirements, High requires a genetic score ≥ 0.5 AND a Layer 1 correlation > 0.3 between tCRE activity and the trait’s cellular profile across cell subtypes.

Medium priority

Meets both minimum requirements but not one or both additional High thresholds. The genetic score is the higher of the SuSiE posterior inclusion probability (PIP) and the Borzoi variant-causality score.

Same-record evidence and aggregation

Each priority requires genetic and cellular evidence supporting the SAME variant–tCRE–gene record. Do not combine thresholds or evidence from different records to create a qualifying link. A gene receives the highest priority observed among its supporting records. A tCRE summary uses supporting records for that exact tCRE, not every trait of its linked genes. Separately displayed aggregate scores do not relax the same-record rule.

Limits of interpretation

A priority is candidate-association evidence, not proof of a causal variant, causal gene, causal cell type or treatment efficacy. Missing evidence is unavailable, not zero. MAGMA −log10(P), displayed credibility scores, cellular-context scores and G/C tier codes must not be substituted for these eligibility tests; use the provided records and their documented fields.