We also integrate scE2G predictions with orthogonal information to prioritize causal genes and cell types for noncoding variants associated with complex traits.

For example, here we nominate regulatory interactions linking INPP4B and IL15 to lymphocyte counts in T cells.

9/12

For example, here we identify cell-type specific links for SPTA1 in erythroblasts and normoblasts.

8/12

We applied scE2G to over 40 cell types from PBMCs, BMMCs, and pancreatic islets, validating that scE2G predictions reflect expected patterns of cell-type specificity.

7/12

Key features in scE2G include 1) ABC score, 2) Kendall correlation between peak accessibility and gene expression, and 3) whether the gene is “ubiquitously-expressed”.

Notably, the Kendall correlation improves long-range predictions and appears to detect stochastic transcriptional bursting.

5/12

In systematic benchmarking against CRISPR perturbations (below), fine-mapped eQTLs, and GWAS variant-gene associations, scE2G models outperforms existing single-cell models and distance-derived baselines.

We applied and extended ENCODE benchmarking pipelines: www.biorxiv.org/content/10.1...

4/12

Using a gold-standard CRISPR perturbation dataset, we trained two logistic regression models: scE2G (ATAC) and scE2G (Multiome), that use single-cell ATAC-seq or single-cell multiomic ATAC and RNA-seq data, respectively.

3/12