This study used an approach called explainable-artificial intelligence (XAI) to identify new biological features. It aimed at understanding the design of guide RNA design and the association of guide RNA with CRISPR-based genome edits. This could improve scientists’ ability to efficiently predict where genomic targets will occur in a genome.
Scientists rely on models to predict where CRISPR-Cas tools act on an organism’s genome. The performance of these models is critically important because these modifications are irreversible. This study by scientists at Oak Ridge National Laboratory and the University of Tennessee, Knoxville aimed to improve the reliability of these tools by using explainable-artificial intelligence to uncover new relationships between the guide RNA, an organism’s DNA, and the activity of CRISPR-based tools. The researchers used publicly accessible datasets to train an explainable artificial intelligence model called iterative Random Forest to predict how efficiently CRISPR-Cas9 can edit specific DNA sequences with a specific guide RNA. Using this approach, the researchers discovered that quantum chemical features had the most significant effect on predicting guide RNA efficiency in both H. sapiens and E. coli. Moreover, the researchers found that the importance of different quantum chemical properties or locations of interest varied with each species. This research underscores the importance of future research in this field to improve the safety and reliability of CRISPR-Cas tools in non-model organisms.
The research was supported by the Secure Ecosystem Engineering and Design project funded by the Genomic Science Program of the Department of Energy Office of Science, Office of Biological and Environmental Research program as part of the Secure Biosystems Design Science Focus Area. One of the researchers was supported by the Center for Bioenergy Innovation, a Department of Energy research center.