Prediction of off-target activities for the end-to-end design of CRISPR guide RNAs

Off-target effects of the CRISPR-Cas9 system can lead to suboptimal gene-editing outcomes and are a bottleneck in its development. Here, we introduce two interdependent machine-learning models for the prediction of off-target effects of CRISPR-Cas9. The approach, which we named Elevation, scores individual guide-target pairs, and also aggregates them into a single, overall summary guide score. We demonstrate that Elevation consistently outperforms competing approaches on both tasks. We also introduce an evaluation method that balances errors between active and inactive guides, thereby encapsulating a range of practical use cases. Because of the large-scale and computational demands of the prediction of off-target activities, we have developed a fast cloud-based service (https://crispr.ml) for end-to-end guide-RNA design. The service makes use of pre-computed on-target and off-target activity prediction for every genic region in the human genome.