scMultiMap: Cell-type-specific mapping of enhancers and target genes from single-cell multimodal data

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Mapping enhancers to target genes in specific cell types is vital for interpreting genome-wide association studies (GWAS) variants.
• Single-cell multimodal data (gene expression, chromatin accessibility) allow cell-type-specific enhancer-gene pairing inference.
• Challenges include data sparsity, variable sequencing depth, and computational demands.

Purpose of the Study:

• Introduce scMultiMap, a novel statistical method for inferring enhancer-gene associations from sparse single-cell multimodal data.
• Address limitations of existing methods, including data sparsity and computational burden.

Main Methods:

• Utilize a joint latent-variable model to infer enhancer-gene associations.
• Incorporate adjustments for technical confounding and provide analytically derived p-values.
• Employ fast moment-based estimation for computational efficiency.

Main Results:

• scMultiMap demonstrates robust type I error control, high statistical power, and significant computational efficiency on blood and brain datasets.
• Achieves computational speed approximately 1% of existing methods.
• Applied to Alzheimer's disease (AD) data, it reveals significant heritability enrichment in microglia.

Conclusions:

• scMultiMap effectively infers cell-type-specific enhancer-gene associations from sparse multimodal data.
• Provides valuable insights into the regulatory mechanisms underlying GWAS variants, particularly for Alzheimer's disease.
• Offers a computationally efficient and statistically powerful approach for single-cell genomics research.