Determinants of base editor efficacy are position-specific. Determinants of base editor efficacy are position-specific

Nucleotide-level control over DNA sequences is poised to power functional genomics studies and lead to new therapeutics. CRISPR/Cas base editors promise to achieve this ability, but the determinants of their activity remain incompletely understood. We measured base editing frequencies in a human cell line for two cytosine base editors at 14,000 target sequences. Base editing activity is sequence-biased, with largest effects from nucleotides flanking the target cytosine, and correlated with measures of Cas9 guide RNA efficiency. The window in which editing occurs depends strongly on the base preceding the cytosine, with a preceding thymine leading to substantial edits at positions 3 to 9 in the target, while a preceding guanine restricts edits to positions 5 to 7. The influence of sequence and guide RNA features on editing rate depends on the position of the edited base within the target. Sequence biases are larger when editing cytosines further from the editing window, while guide RNA features have larger bias in the centre of the editing window. We use these observations to train a machine learning model to predict editing activity at every targetable cytosine, with accuracy ranging from 0.55 to 0.75 across positions. We demonstrate the usefulness of our model by predicting the efficacy of potential disease mutation correcting guides, and find that most of them suffer from more unwanted editing than corrected outcomes. This work highlights the position-specificity of base editing biases, and provides a solution to account for them, thus allowing more efficient planning of base edits in experimental and therapeutic contexts.