Nicole F. Oldham Green, Gavin J. Sutton, Javier Pérez‐Burillo, J. Wang, Samuel Bagot, Hannah G Danon, Kieran Walsh, Akira Gokoolparsadh, Samantha A. Miles, Guang Yang, Charles A. Herring, Yuheng Liang, Grant Pfundstein, Vladimir Sytnyk, Hamid Alinejad-Rokny, Ryan Lister, Joseph Rosenbluh, Johann A. Gagnon-Bartsch, Irina Voineagu
Genetic variants associated with complex traits often lie in distal enhancers. While candidate enhancers have been mapped genome wide, their functional state and gene targets in specific cell types remain unclear. Here we present AstroREG, a resource of enhancer-gene interactions in human primary astrocytes, generated by combining CRISPR inhibition (CRISPRi), single-cell RNA-seq and machine learning. By functionally testing nearly 1,000 PsychENCODE enhancers, we identified more than 150 regulatory interactions, revealing enhancers that control key astrocyte functions and genes implicated in Alzheimer's disease. The CRISPRi screen also provided valuable ground-truth data from a primary cell type for training and benchmarking prediction models of enhancer activity. We thus developed EGrf, a random forest (RF) model trained on these data, and applied it genome wide to predict regulatory interactions with high specificity. Together, our data provide a comprehensive functional map of enhancer-mediated regulation in a key glial cell type, shedding light on brain function and disease.