Connecting polygenic disease risk to cell states and regulatory programs through single-cell chromatin accessibility

The paper introduces SCADS, a scalable computational framework that integrates single-cell ATAC-seq data with GWAS summary statistics to identify disease-relevant cell populations and regulatory programs by quantifying the polygenic enrichment of chromatin co-regulatory regions.

The Gist

Genetic studies have long identified thousands of small variations in our DNA that are associated with an increased risk of developing diseases. Most of these variations do not sit inside the genes that provide instructions for building proteins. Instead, they sit in the regulatory regions of the genome. These regions act like switches that control when, where, and how much a gene is turned on or off. Because these switches can be specific to certain types of cells, a genetic variation might increase disease risk in a lung cell but have no effect in a skin cell.

The researchers in this paper developed a computational framework called Single-Cell ATAC-seq Disease Score (SCADS) to bridge this gap. The goal is to connect these genetic risk signals to specific cell types by looking at which parts of the DNA are accessible for regulation in individual cells.

To do this, the researchers used a technique called single-cell ATAC-sequencing. This method identifies which parts of the DNA are "open" or accessible in a single cell, indicating that those regions are active regulatory switches. The SCADS framework works in three steps. First, it uses a mathematical model to group these open DNA regions into "topics." These topics represent sets of regulatory elements that work together to control specific biological programs. Second, the framework calculates how much the known genetic risk signals for a specific disease are concentrated within these topics. Third, it combines the activity of these topics within individual cells to produce a disease score for every single cell.

The authors tested SCADS using simulations and found that it identifies disease-relevant cells with higher accuracy and fewer false positives than existing methods. They also demonstrated that the scores are calibrated, meaning they can be compared across different datasets and different diseases.

When the researchers applied SCADS to autoimmune diseases, they found that disease relevance is not uniform across all cells of a single type. For example, in inflammatory bowel disease, the researchers found that different subsets of CD8+ T cells and different types of colon epithelial cells carried different levels of disease risk. By using SCADS, the researchers were able to pinpoint the specific gene programs and genetic variants that drive these differences. The paper describes SCADS as a scalable and modular framework for connecting noncoding genetic variation to the identity and function of individual cells.

Technical Summary

Technical Summary: Connecting Polygenic Disease Risk to Cell States and Regulatory Programs through Single-Cell Chromatin Accessibility

1. Problem Statement

Genome-wide association studies (GWAS) have identified thousands of genetic variants associated with complex polygenic diseases. However, the vast majority of these variants reside in non-coding regions, making it difficult to determine their functional impact. While single-cell ATAC-sequencing (scATAC-seq) provides high-resolution maps of chromatin accessibility, linking these regulatory landscapes to specific disease risks remains a challenge.

Existing methods often struggle to:

• Account for cell-type heterogeneity: Disease risk may be driven by specific sub-populations rather than entire cell types.
• Handle data sparsity and noise: scATAC-seq data is inherently sparse.
• Ensure comparability: Many methods produce scores that are difficult to compare across different datasets or different disease traits due to lack of calibration.

2. Methodology: The SCADS Framework

The authors propose SCADS (Single-Cell ATAC-seq Disease Score), a computational framework designed to integrate GWAS summary statistics with scATAC-seq data. The methodology follows a three-step pipeline:

1. Topic Modeling for Co-regulatory Regions: Instead of analyzing individual peaks (which are sparse and noisy), SCADS uses topic modeling to identify "chromatin co-regulatory regions." These are groups of genomic elements that exhibit similar accessibility patterns across cells, effectively capturing functional regulatory modules or "topics."
2. Quantifying Polygenic Enrichment: For each identified topic, the framework calculates the enrichment of GWAS signals. It determines whether the genomic regions associated with a specific topic are significantly more likely to contain disease-associated variants compared to the genomic background.
3. Calibrated Cell-Level Scoring: SCADS calculates a specific disease score for every individual cell. This score is a function of the cell's "topic weights" (how much each regulatory module is active in that cell) and the "enrichment levels" of those topics. Crucially, the scores are calibrated, allowing researchers to compare disease relevance across different cell types, different datasets, and different diseases.

3. Key Contributions

• A Novel Integration Framework: SCADS bridges the gap between population-level genetic associations (GWAS) and single-cell epigenetic landscapes (scATAC-seq).
• Dimensionality Reduction via Topic Modeling: By shifting the unit of analysis from individual peaks to co-regulatory topics, the method increases statistical power and biological interpretability.
• Scalability and Modularity: The framework is designed to be computationally efficient and can be applied to various single-cell modalities and diverse disease traits.
• Calibration: Unlike previous methods, SCADS provides scores that are standardized, enabling cross-study comparisons.

4. Results

• Benchmarking and Simulation: Extensive simulations demonstrated that SCADS outperforms existing state-of-the-art methods in terms of statistical power while maintaining strict control over false-positive rates.
• Biological Application (Autoimmune Traits): When applied to autoimmune diseases, SCADS revealed that disease relevance is not uniform across canonical immune cell types. It identified significant heterogeneity within cells that were previously thought to be homogeneous in their disease contribution.
• Case Study (Inflammatory Bowel Disease - IBD):
  • CD8+ T cells: SCADS pinpointed specific sub-populations of CD8+ T cells that drive IBD risk.
  • Colon Epithelial Cells: The method identified specific epithelial cell states and underlying gene regulatory programs linked to IBD.
  • Variant Mapping: The framework successfully pinpointed the specific genetic variants and regulatory programs driving these cellular phenotypes.

5. Significance

SCADS represents a significant advancement in functional genomics. By moving beyond "cell-type-level" analysis to "cell-state-level" analysis, it allows researchers to pinpoint the exact cellular contexts where genetic risk is manifested. This provides a powerful tool for:

• Mechanistic Discovery: Identifying the specific gene programs and regulatory elements that mediate disease risk.
• Precision Medicine: Understanding how different patient populations might have different cellular drivers for the same disease.
• Drug Target Identification: Highlighting specific cell states and regulatory modules that could be targeted therapeutically to mitigate disease risk.