Prime editing is a ‘search-and-replace’ genome-editing tool2. The system consists of a prime editor protein — a nicking Cas9 fused to reverse transcriptase — that takes direction from the information encoded in a pegRNA. Within each pegRNA is a protospacer (the ‘search’ sequence) that directs the prime editor to a site of interest, as well as a 3’ extension (the ‘replace’ sequence) that provides the template for reverse transcriptase to synthesize a DNA strand containing the desired mutation.

A key obstacle in prime editing is deciding among the hundreds of possible pegRNA designs. For any desired mutation, many design combinations are possible through variation in the protospacer and the two sub-components of the 3’ extension: the so-called primer binding site, and the reverse transcriptase template. Most pegRNAs will not edit efficiently, and searching the vast pegRNA design space to identify efficient pegRNAs is challenging. Machine learning methods that predict pegRNA efficiency continue to progress3,4, but even these algorithms are not guaranteed to yield a high-efficiency pegRNA.