Beyond Genomics: Single-Cell Genomics

Single-cell genomics (scRNA-seq, scATAC-seq, spatial transcriptomics, and multiome assays) resolves molecular variation at the level of individual cells. When integrated with GWAS, these data enable a conceptual shift from "locus discovery" to "cellular mechanism inference", linking genetic risk to specific cell types, states, regulatory programs, and spatial contexts.

This section presents a refined framework for GWAS–single-cell integration, organized by the biological question being asked and the resolution of inference.

Why Integrate GWAS with Single-Cell Data?

GWAS robustly identifies trait-associated loci, but typically leaves some key questions unanswered:

1. Which cell types mediate genetic risk?
2. Which cellular states or programs are involved?
3. Where in tissue architecture does risk manifest?

Single-cell datasets address these gaps by enabling:

• Cell-type and cell-state resolution of GWAS signals
• Gene and regulatory program prioritization in relevant cells
• Dissection of heterogeneous tissues (immune system, brain, tumor microenvironment)
• Cell-type-specific genetic architectures (e.g. sc-eQTL, scATAC-QTL, state-dependent effects)

Framework for GWAS–Single-Cell Methods

Methods can be organized along two orthogonal axes:

Genetic abstraction level

• Variant / heritability-based
• Gene-based
• Cell-based
• Spatial / tissue-context-based

Biological resolution

• Cell type
• Cell state / program
• Regulatory mechanism
• Spatial niche

Approaches

1. Cell-type heritability and gene-set enrichment

(Variant-level or gene-level; population-wide signal)

Representative methods: - LDSC-seg (stratified LD Score Regression) - GitHub - MAGMA (gene-property and gene-set analysis) - Software

Core question: "Are GWAS signals enriched in genes or annotations specific to certain cell types?"

Core idea: - Derive cell-type-specific annotations from expression or chromatin data - Partition GWAS heritability (LDSC-seg) or gene-level association (MAGMA) - Test enrichment while accounting for LD and baseline genomic features

Typical inputs: - GWAS summary statistics - Cell-type aggregated expression or accessibility profiles - LD reference panels

Typical outputs: - Enrichment statistics per cell type or tissue

Strengths: - Robust, LD-aware, interpretable - Ideal as a first-pass prioritization step

Limitations: - Limited resolution (cell type rather than individual cells) - Sensitive to gene-to-SNP mapping choices

Workflow:

GWAS Summary Statistics
         ↓
    LD Reference Panels
         ↓
Cell-type Aggregated Expression/Accessibility Profiles
         ↓
    [LDSC-seg: Stratified LD Score Regression]
    [MAGMA: Gene-property Analysis]
         ↓
    Enrichment Statistics per Cell Type

2. Per-cell and per-state disease relevance scoring

(Cell-level resolution; heterogeneity-aware)

Representative methods: - scDRS - GitHub

Core question: "Which individual cells or cellular states are most relevant to a given disease?"

Core idea: - Convert GWAS summary statistics into gene-level disease scores - Score each cell by coordinated expression of disease-associated genes - Use matched control gene sets for calibration and statistical testing

Typical inputs: - scRNA-seq data (cell-by-gene expression matrix) - GWAS summary statistics or derived gene scores

Typical outputs: - Disease relevance score per cell - Cluster- or state-level summaries

Strengths: - Captures within–cell-type heterogeneity - Highlights rare, transient, or activated states

Limitations: - Expression-based (does not directly model regulatory variants) - Interpretation is correlational rather than causal

Workflow:

GWAS Summary Statistics
         ↓
    Gene-level Disease Scores
         ↓
    scRNA-seq Data (Cell × Gene Matrix)
         ↓
    [scDRS: Score Each Cell]
         ↓
    Disease Relevance Score per Cell
         ↓
    Cluster/State-level Summaries

3. Variant-to-gene-to-cell-type linking

(Mechanistic and regulatory interpretation)

Representative methods: - sc-linker - GitHub

Core question: "Which genes mediate GWAS loci, and in which cell types do they act?"

Core idea: - Integrate GWAS loci with single-cell chromatin accessibility and expression - Link non-coding variants to regulatory elements - Connect regulatory elements to target genes in a cell-type-specific manner

Typical inputs: - GWAS summary statistics or fine-mapped loci - scATAC-seq / multiome data - Single-cell gene expression

Typical outputs: - Prioritized causal genes per locus - Cell-type-specific regulatory links

Strengths: - Moves toward causal interpretation - Explicitly models regulatory context

Limitations: - Requires high-quality regulatory maps - Peak-to-gene linking remains noisy

Workflow:

GWAS Summary Statistics / Fine-mapped Loci
         ↓
    scATAC-seq / Multiome Data
         ↓
    Single-cell Gene Expression
         ↓
    [sc-linker: Link Variants → Regulatory Elements → Genes]
         ↓
    Prioritized Causal Genes per Locus
         ↓
    Cell-type-specific Regulatory Links

4. Spatial mapping of genetic risk

(Tissue architecture and microenvironment context)

Representative methods: - gsMap - GitHub

Core question: "Where in the tissue does genetic risk manifest?"

Core idea: - Use Graph Neural Network (GNN) to identify homogeneous spots (microdomains) based on gene expression patterns and spatial coordinates - Compute gene specificity scores (GSS) for each spot by comparing gene expression ranks within microdomains versus the entire section - Map GSS to SNPs via distance-based windows (±50 kb from transcription start sites) and SNP-to-gene linking maps - Apply stratified LD Score Regression (S-LDSC) to test whether SNPs with higher GSS are enriched for trait heritability - Aggregate spot-level associations to spatial regions using the Cauchy combination test

Typical inputs: - Spatial transcriptomics data (with spatial coordinates and gene expression profiles) - GWAS summary statistics - LD reference panels - SNP-to-gene linking maps (e.g., from epigenomic data) - Optional: cell type annotation priors

Typical outputs: - Enrichment statistics and P-values per spatial spot - Spatial maps of trait-associated cells or regions - Region-level aggregated associations

Strengths: - Addresses sparsity and technical noise in ST data through microdomain aggregation - Provides spatially resolved mapping at cellular resolution - Adds anatomical and microenvironmental context - Essential for brain, cancer, and developmental studies

Limitations: - Resolution depends on spatial technology (spot-level in high-resolution platforms, cluster-level in conventional platforms) - Requires high-quality SNP-to-gene linking maps - Computational intensity increases with spatial resolution

Workflow:

Spatial Transcriptomics Data (Expression + Coordinates)
         ↓
    [GNN: Identify Homogeneous Spots / Microdomains]
         ↓
    [Compute Gene Specificity Scores (GSS) per Spot]
         ↓
    [Map GSS to SNPs via Distance & SNP-to-Gene Links]
         ↓
    GWAS Summary Statistics + LD Reference Panels
         ↓
    [S-LDSC: Test Heritability Enrichment per Spot]
         ↓
    [Cauchy Combination Test: Aggregate to Regions]
         ↓
    Spatial Maps of Trait-associated Spots/Regions

Conceptual Summary Table

References

Review papers

• Cuomo, A. S. E., Nathan, A., Raychaudhuri, S., MacArthur, D. G., & Powell, J. E. (2023). Single-cell genomics meets human genetics. Nature Reviews Genetics, 24(8), 535–549. https://doi.org/10.1038/s41576-023-00598-6

Method papers

• Finucane, H. K., Reshef, Y. A., Anttila, V., Slowikowski, K., Gusev, A., Byrnes, A., et al. (2018). Heritability enrichment of specifically expressed genes identifies disease-relevant tissues and cell types. Nature Genetics, 50(4), 621–629. https://doi.org/10.1038/s41588-018-0081-4

• de Leeuw, C. A., Mooij, J. M., Heskes, T., & Posthuma, D. (2015). MAGMA: generalized gene-set analysis of GWAS data. PLoS Computational Biology, 11(4), e1004219. https://doi.org/10.1371/journal.pcbi.1004219

• Zhang, M. J., Hou, K., Dey, K. K., et al. (2022). Polygenic enrichment distinguishes disease associations of individual cells in single-cell RNA-seq data. Nature Genetics, 54(9), 1344–1350. https://doi.org/10.1038/s41588-022-01167-z (scDRS)

• Jagadeesh, K. A., Dey, K. K., Montoro, D. T., Mohan, R., Gazal, S., Engreitz, J. M., Xavier, R. J., Price, A. L., Regev, A., et al. (2022). Identifying disease-critical cell types and cellular processes by integrating single-cell RNA-sequencing and human genetics. Nature Genetics, 54(10), 1479–1492. https://doi.org/10.1038/s41588-022-01187-9 (sc-linker)

• Song, L., Chen, W., Hou, J., Guo, M., & Yang, J. (2025). Spatially resolved mapping of cells associated with human complex traits. Nature, 641, 932–941. https://doi.org/10.1038/s41586-025-08757-x (gsMap)