Scalable variant-to-function screens using prime editing perturb-seq

Linking genetic variants to their functional consequences at scale remains a major challenge in human genetics. Perturb-seq is limited in the variants it can install, while prime editing has largely been restricted to simpler mechanistic readouts. Here, we introduce prime editing perturb-seq (PEP-seq), which links programmed mutations to transcriptomic states. We designed 2,164 prime editing guides targeting common expression quantitative trait loci (eQTLs), premature termination codons, and position-matched synonymous controls, generating 749,110 time-resolved single-cell transcriptomes. PEP-seq precisely detected nonsense-mediated decay, while probe-based junction measurements suggested depletion of mature RNAs. For eQTLs, a locus-size-corrected specificity metric showed that the intended target was frequently the most affected transcript within a 1 Mb window, with effects also associated with open chromatin near the edit site. These results establish PEP-seq as a scalable platform for direct, endogenous variant-to-function mapping.