Mapping enhancer-gene regulatory interactions from single-cell data.

Family of classification models (scE2G) using features from single-cell ATAC-seq or multiomic RNA and ATAC-seq data, trained on a CRISPR perturbation dataset to predict enhancer-gene pairs.

Linking an enhancer to its target gene in a given cell type underpins the interpretation of noncoding variants, yet predicting these interactions from single-cell data has remained difficult. The authors introduce scE2G, a family of classification models using features from single-cell ATAC-seq or multiomic RNA and ATAC-seq data, trained on a CRISPR perturbation dataset covering more than 10,000 evaluated element-gene pairs. The models were benchmarked against CRISPR perturbations, fine-mapped expression quantitative trait loci and genome-wide association study variant-gene associations, with state-of-the-art performance across several cell types and categories of perturbation. Applied to heterogeneous tissues, scE2G produced maps of enhancer-gene regulatory interactions and nominated regulatory interactions linking INPP4B and IL15 to lymphocyte count.

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

This sits one step upstream of diagnosis: nothing here plugs directly into the interpretation of a noncoding variant in a patient, since the demonstration involves a complex trait and public functional data. The medium-term value is real, because a genome with no causal coding variant leaves regulatory sequence as the remaining avenue and tissue-specific enhancer-gene maps are exactly what we lack. One to follow for teams building noncoding annotation for their genomes, not for the routine laboratory.

Analysis by Dr Thibaut Benquey