A research group led by DBCLS Visiting Professor Hidemasa Bono, Visiting Associate Professor Yuki Naito, and former member Hiromasa Ono (now at PtBio Inc.) has published a paper entitled “Risk Prediction of RNA Off-Targets of CRISPR Base Editors in Tissue-Specific Transcriptomes Using Language Models. The paper was published in the International Journal of Molecular Sciences (IJMS), Special issue: Research Advances in the Bioinformatics of Genome Editing and Gene Function Analysis. The article can be found at the following URL

https://doi.org/10.3390/ijms26041723

Base editing technology (particularly cytosine base editor: CBE) can modify genes without DNA double-strand breaks, however, unintended RNA editing (RNA off-target effects) may occur, and the actual situation has not been fully clarified.
In this study, newly developed CBE-induced RNA off-target effect analysis pipeline, PiCTURE has revealed that unexpected edits occur in addition to previously known patterns of RNA editing.
In addition, based on the DNABERT-2, one of the DNA language models, machine learning models STL and SNL, also called RNAOffScan, were developed to predict RNA off-target risk with a high accuracy. The PROTECTiO pipeline was newly established to demonstrate the practical application of the predictive model for CBE-induced RNA off-target risk. The pipeline estimated risk in specific tissues, for example, low risk in the brain and ovaries, but high risk in the colon and lungs.
This study offers a standard method for assessing RNA off-target risks, marking a significant step toward improving the safety of CBE and is expected to contribute to base-editing specificity and therapeutic applicability.