Graph-CRISPR: a gene editing efficiency prediction model based on graph neural network with integrated sequence and...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• CRISPR gene-editing technology revolutionized molecular biology.
• Accurate prediction of gene-editing efficiency is vital for optimizing applications.
• Existing computational models often lack generalizability across diverse CRISPR systems and conditions, frequently overlooking RNA secondary structure.

Purpose of the Study:

• To develop a novel graph-based computational model, Graph-CRISPR, for predicting gene-editing efficiency.
• To integrate both single-guide RNA (sgRNA) sequence and secondary structure features into the predictive model.
• To enhance the accuracy and adaptability of gene-editing efficiency predictions across various CRISPR systems.

Main Methods:

• Developed Graph-CRISPR, the first graph-based model incorporating sgRNA sequence and secondary structure information.
• Utilized graph neural networks to process and integrate diverse feature types.
• Evaluated model performance against baseline models across different CRISPR editing systems (e.g., CRISPR-Cas9, prime editing, base editing).

Main Results:

• Graph-CRISPR demonstrated superior performance compared to existing baseline models across multiple gene-editing systems.
• The model exhibited strong resilience and maintained robust predictive accuracy under varying experimental conditions.
• Integration of secondary structure features significantly improved prediction accuracy.

Conclusions:

• Graph-CRISPR represents a significant advancement in predicting gene-editing efficiency by leveraging sequence and structural data.
• Graph-based modeling offers a powerful approach to enhance the accuracy and adaptability of computational tools for gene editing.
• The findings underscore the importance of incorporating structural information for more effective CRISPR technology optimization.