Predicting gene-specific regulation with transcriptomic and epigenetic single-cell data.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Single-cell (sc) ATAC-seq and scRNA-seq data provide deep insights into gene regulation.
• Analyzing sparse sc data for gene regulation remains challenging.

Purpose of the Study:

• To develop a novel computational approach, MetaFR, for modeling gene expression from single-cell chromatin accessibility and gene expression data.
• To assess the performance and efficiency of MetaFR compared to existing state-of-the-art methods.

Main Methods:

• MetaFR employs regression trees to build gene-specific models linking scATAC-seq and scRNA-seq data.
• Models can be trained at the single-cell or meta-cell level.
• Model performance is validated using fine-mapped expression quantitative trait loci (eQTLs).

Main Results:

• MetaFR accurately predicts gene expression by linking open-chromatin variation to gene expression.
• Meta-cell models generally outperform single-cell models.
• MetaFR demonstrates superior runtime and prediction performance compared to the SOTA method SCARlink.

Conclusions:

• MetaFR offers a time-efficient and reliable method for modeling gene expression.
• The approach enables the study of gene regulation across various organisms with available scRNA-seq and scATAC-seq data.