Genetic enrichment

When to use

Use this module to ask which cell types or nuclei have accessible chromatin overlapping genetic evidence for a human trait. It weights ATAC peaks by the fine-mapping probabilities of the GWAS variants they contain, scores enrichment per cell type and per nucleus, and uses the SCAVENGE method to propagate nucleus scores through the WNN neighbor graph. Stratified LD score regression (S-LDSC) adds a genome-wide test: whether each cell type’s specifically accessible peaks, or specifically expressed genes, carry more of a trait’s heritability than peaks or genes in general. Method details

A credible set contains candidate causal variants at a GWAS locus, with probabilities from fine-mapping. Enrichment helps prioritize cellular contexts; it does not by itself identify a causal cell type, gene or mechanism.

The public demo leaves this module off. Its configuration already contains a study list for immune_human_2x, shown under Parameter reference; add genetic_enrichment to that aggregation’s modules to try it.

Prerequisites

Before enabling the module, confirm that:

the aggregation is human and you have reviewed it through checkpoint 8, including its cell-type annotations;
each configured study represents the intended trait and population; S-LDSC uses European LD, so it only approximates studies of other ancestries;
the machine can download and keep about 4 GB of Open Targets Parquet data, stored once under <store>/files/OpenTargets/;
for S-LDSC, the machine can download a 1.5 GB LD reference from Zenodo, kept as 2.8 GB under <store>/files/S-LDSC/, and each study’s genome-wide summary statistics from the GWAS Catalog; and
you have decided whether to interpret cell-type summaries, individual nuclei or graph-propagated scores; these answer related but different questions.

Outputs

Question	Output
Which variants and peaks carry each study’s evidence?	Peak-weight summaries and study and variant-to-peak tables
Do the studies share fine-mapped variants?	A heatmap of credible-set similarity between studies
Which cell types are enriched, and which matters most?	Cell-type chromVAR deviation heatmaps, which outline each study’s top set of cell types
Which loci and variants drive a cell type’s enrichment?	Locus-contribution heatmaps and bars, and variant detail plots
Which nuclei are trait-relevant?	SCAVENGE trait-relevance scores on the WNN graph, summarized by cell type and cluster
Is a study’s genome-wide heritability enriched in a cell type’s specific peaks or genes?	S-LDSC heatmap, which outlines each study’s top set of cell types, and a comparison of the peak, gene and joint models
Do S-LDSC and chromVAR agree?	S-LDSC versus chromVAR scatter plots per study and an agreement overview
How robust is an S-LDSC enrichment, and do peaks or genes explain more?	A plot of each enrichment’s most influential genome blocks, and comparisons of peak and gene annotations of matched size

Configure

Add genetic_enrichment to the aggregation’s modules list, keeping its other settings:

cfg_aggregations.yaml

my_aggregation:
  modules: [genetic_enrichment]

Then add an entry for my_aggregation to cfg_module_genetic_enrichment.yaml in the selected configuration directory:

cfg_module_genetic_enrichment.yaml

my_aggregation:
  genetic_enrichment_GWAS_studies:
    lymphocyte_count:
      Category: positive_control
      sourceId: GCST90002388
      finemappingMethod: auto

Each entry under genetic_enrichment_GWAS_studies is one study. Its name labels the results, and Category groups studies in the plots.

A sourceId names an Open Targets GWAS study with credible sets, read from the Open Targets release 26.03 pinned by the workflow: a GWAS Catalog accession such as GCST90002388, or a FinnGen study such as FINNGEN_R12_T2D. The run fails if the release has no credible sets for the study. finemappingMethod: auto, the default, selects the first available method in the order SuSie, SuSiE-inf, PICS. Name a method instead when the method is part of the analysis; the run then fails if that method is unavailable for the study.

A sourceId ending in .parquet is instead the path of a local file, relative to the repository root. Omit finemappingMethod for local files: the method, like the study ID, genome build and provenance, is read from the file.

S-LDSC needs genome-wide summary statistics rather than credible sets. For a GWAS Catalog sourceId, they are fetched from the GWAS Catalog, which keeps harmonised files for many but not all studies. Name other summary statistics with summaryStatistics: another GWAS Catalog accession, or the path or URL of a GRCh38 GWAS-SSF table with rsIDs, such as a consortium release:

cfg_module_genetic_enrichment.yaml

my_aggregation:
  genetic_enrichment_GWAS_studies:
    type_2_diabetes:
      Category: metabolic
      sourceId: FINNGEN_R12_T2D
      summaryStatistics: GWAS/T2D_consortium_GRCh38.tsv.gz

The parameter browser lists the columns that S-LDSC reads. Studies without summary statistics, such as FinnGen studies or local files without summaryStatistics, appear as grey rows in the S-LDSC plots. The other analyses use their credible sets as before.

Columns of a local fine-mapped GWAS file

The file holds one GWAS on GRCh38, with one row per credible-set variant. Its columns follow the Open Targets credible-set dataset, with the locus array flattened into one row per variant, and that dataset documents their meaning. multiomeR uses these columns:

Column	Content	Used for
`studyId`, `finemappingMethod`, `confidence`, `sampleSize`, `sourceRelease`	One value per file	Study labels and metadata
`credibleSetProbability`	One value, greater than 0 and at most 1	Study metadata
`studyLocusId`, `credibleSetIndex`	Credible-set identifiers	Grouping variants into loci
`variantId`	`chromosome_position_REF_ALT`	Matching variants between studies
`chromosome`, `position`	GRCh38 coordinates, chromosome without `chr`	Overlap with peaks
`posteriorProbability`	0 to 1	Peak weights
`beta`	Effect size, or missing	Effect-weighted enrichment
`locusStart`, `locusEnd`	Bounds that contain each variant’s `position`	Locus plots

The file must also contain schemaVersion (1), studyType (gwas), genomeBuild (GRCh38), sourceUrl, sourceSha256 (lowercase SHA-256) and sourcePublication, each with one value, and the columns variantRepresentation, logBF, pValueMantissa, pValueExponent, standardError, r2Overall, is95CredibleSet and is99CredibleSet. Each studyId, studyLocusId and variantId combination must be unique. A file with credibleSetProbability 0.95 or 0.99 may contain only variants in that set.

Run

Preview the targets tagged for this module:

targets::tar_manifest(
  names = targets::tar_described_as(
    tidyselect::contains("checkpoint:genetic_enrichment")
  ) & tidyselect::ends_with(".my_aggregation"),
  callr_function = NULL
)[, c("name", "description")]

Then run the same selection:

targets::tar_make(
  names = targets::tar_described_as(
    tidyselect::contains("checkpoint:genetic_enrichment")
  ) & tidyselect::ends_with(".my_aggregation")
)

Runtime and disk use grow with the number of studies, nuclei and attributed loci. S-LDSC downloads each study’s summary statistics once for all aggregations, one study at a time to spare the GWAS Catalog’s server, which takes about a minute per study.

Review

Review the plots (examples); each plot’s subtitle and caption say what to look for.

<store>/plots/my_aggregation/genetic_enrichment/
├── GWAS_peak_weights_barplot.png
├── GWAS_credible_set_similarity_heatmap.png
├── cell_type_pseudobulk/
│   ├── chromVAR_deviation_heatmaps/
│   ├── chromVAR_locus_contribution_per_GWAS_heatmaps/
│   ├── chromVAR_locus_contribution_per_GWAS_faceted_bars_plots/
│   ├── chromVAR_absolute_effect_locus_contribution_per_GWAS_faceted_bars_plots/
│   ├── chromVAR_variant_contribution_detail_plots/
│   ├── SLDSC_heatmap.png
│   ├── SLDSC_modality_comparison_plot.png
│   ├── SLDSC_block_influence_plot.png
│   ├── SLDSC_chromVAR_comparison_plot.png
│   ├── SLDSC_chromVAR_agreement_plot.png
│   ├── SLDSC_matched_enrichment_curves_plot.png
│   ├── SLDSC_matched_size_comparison_plot.png
│   └── SLDSC_matched_enrichment_agreement_plot.png
└── single_nucleus/SCAVENGE/WNN_harmony_SNN/
    ├── TRS_heatmap/
    ├── TRS_UMAPs/
    └── TRS_cluster_summary_plot/

Read the tables behind the plots in R, for example the cell-type enrichment scores and the S-LDSC results:

targets::tar_read(
  chromVAR_deviation_tibble.cell_type_pseudobulk.genetic_enrichment.my_aggregation
)
targets::tar_read(
  SLDSC_cell_type_tibble.cell_type_pseudobulk.genetic_enrichment.my_aggregation
)

Parameter reference

Open the searchable parameter browser. Genetic-enrichment methods describes the analyses and their key fixed values, links to the source files and shows the target structure.

Show the public immune_human_2x example

YAML

immune_human_2x:
  genetic_enrichment_GWAS_studies:
    LymphocyteCount_Vuckovic2020:
      Category: positive_control
      sourceId: GCST90002388
    MonocyteCount_Vuckovic2020:
      Category: positive_control
      sourceId: GCST90002393
    BCellNeoplasm_Guler2025:
      Category: B_cell_neoplasm
      sourceId: GCST90624736
    BMI_Verma2024:
      Category: negative_control
      sourceId: GCST90475159
    EducationalAttainment_Okbay2022:
      Category: brain_cognition
      sourceId: GCST90105038