Prime editing is a versatile genome editing technology that enables precise genetic modifications without inducing DNA double-strand breaks. Owing to numerous variables in the prime editing guide RNA (pegRNA) design, experimentally identifying the most efficient pegRNA for a specific locus and edit is laborious. Therefore, we have developed computational tools to streamline this process. Here we present a comprehensive protocol detailing how to use PRIDICT2.0 and ePRIDICT, machine-learning models that assess the influence of the pegRNA design and chromatin context on prime editing. PRIDICT2.0 is an ensemble of attention-based bidirectional recurrent neural networks that predicts pegRNA efficiencies for replacements, insertions or deletions in different cellular contexts. Compared with other pegRNA design tools, PRIDICT2.0 accommodates larger edits-up to 40 base pairs-across diverse edit types, also allowing the introduction of silent bystander edits that can enhance editing efficiency. ePRIDICT, a gradient-boosting algorithm, further accounts for the local chromatin environments and assesses how the genomic location of the target site affects prime editing rates. Both tools are available at www.pridict.it for individual predictions or can be installed locally for batch processing of multiple edits and target sites. The protocol provides step-by-step instructions on using PRIDICT2.0 and ePRIDICT, covering sequence input, prediction generation and interpretation. Web-based predictions take under a minute, while local installation and batch processing may take up to several hours, depending on the dataset size. By streamlining pegRNA selection and chromatin context analysis, these tools promote the adoption of prime editing in basic and translational research.
Mathis et al. (Thu,) studied this question.