Evaluation of computational tools for predicting CRISPR gRNA on-target efficiency in plants.

CRISPR technologies has become an integral part of plant biotechnology, synthetic biology and basic plant research, routinely used by researchers for targeted genome modifications. CRISPR guide RNAs (gRNAs) undermines the highly programmable nature of CRISPR, enabling site-specific genome editing. However, different gRNA targets showed highly variable on-target effectiveness and poor gRNA design could amount to wasting valuable scientific resources. There has been broad development of computational and web-based tools for gRNA efficiency predictions but their performances in plant genome editing remains controversial or untested. Hence, in this study, we systematically evaluated over 20 accessible, web-based in silico gRNA on-target efficiency prediction tools using an experimental plant genome editing dataset. Excitingly, we identified multiple tools, mostly developed using machine learning, that were highly predictive of gRNA on-target genome editing efficiency in planta. The prediction scores assigned to gRNAs in the dataset by these tools were significantly correlated with the frequency of CRISPR-mediated InDels in plants. Furthermore, we evaluated efficiency prediction scores available on popular platforms such as CRISPOR and CRISPR-P which contain large numbers of non-model plant genomes. Our analysis showed that some prediction scores on CRISPOR performed quite well which allows efficient integration of on-target and off-target predictions. Overall, we believe that our study provided insights on improving gRNA design during conventional plant genome editing workflows and should also help unfamiliar researchers interested in CRISPR/SpCas9 genome editing.