Distinguishing causal from tagging enhancers using single-cell multiome data.

Area of Science:

• Genomics
• Epigenetics
• Computational Biology

Background:

• Single-cell RNA-seq+ATAC-seq multiome data links enhancers to genes via chromatin accessibility and gene expression.
• Correlations among ATAC-seq peaks can lead to non-causal "tagging" peak-gene links.

Purpose of the Study:

• To investigate and quantify the pervasive "tagging" effects in peak-gene linking using multiome data.
• To develop methods to distinguish causal from non-causal enhancer-gene associations.

Main Methods:

• Defined "co-accessibility" and "co-activity" scores for ATAC-seq peaks based on correlations with nearby peaks and genes.
• Analyzed four multiome datasets (86k cells, 6 immune/blood cell types).
• Utilized CRISPRi and eQTL data for validation and regression analysis (S-CASC).

Main Results:

• Co-accessibility and co-activity scores were highly correlated (r=0.57-0.73), indicating pervasive tagging effects.
• Tagging effects were not explained by technical factors but were consistent with co-accessibility.
• Causal associations were enriched in specific functional regions (e.g., near TSS, H3K27ac marks) and driven by pioneer transcription factors.
• Fine-mapped peak-gene links using SuSiE outperformed marginal links in validation datasets.

Conclusions:

• Tagging effects due to peak co-accessibility are widespread in single-cell multiome data, complicating enhancer-gene linking.
• Distinguishing causal from non-causal links requires accounting for these tagging effects.
• Accurate enhancer-gene association is vital for understanding gene regulation and interpreting GWAS findings in complex traits.