Learning to utilize internal protein 3D nanoenvironment descriptors in predicting CRISPR-Cas9 off-target activity

Area of Science:

• Biochemistry
• Molecular Biology
• Bioinformatics

Background:

• CRISPR-Cas9 gene editing relies on single guide RNA (sgRNA) binding to target DNA.
• Predicting off-target cleavage by CRISPR-Cas9 remains challenging due to complex interactions.
• Physico-chemical and structural descriptors of sgRNA-DNA interactions are not fully understood.

Purpose of the Study:

• To comprehensively assess physico-chemical/structural descriptors influencing CRISPR-Cas9 cleavage activity.
• To exploit the protein 3D nanoenvironment of sgRNA-(off-)target DNA pairs for predictive modeling.
• To develop a machine learning model for predicting CRISPR-Cas9 off-target cleavage.

Main Methods:

• Harvested 634,980 residue-level features from CRISPR-Cas9 complexes.
• Simulated the internal protein 3D nanoenvironment for single-base mutations.
• Developed and validated the STING_CRISPR machine learning model.

Main Results:

• Identified key residue-level features influencing CRISPR-Cas9 off-target cleavage.
• STING_CRISPR demonstrated accurate predictive performance for single-base mutations.
• Discovered four critical Cas9 residue spatial hotspots affecting cleavage activity.

Conclusions:

• The study provides a novel approach to understanding CRISPR-Cas9 off-target effects.
• STING_CRISPR offers a valuable tool for enhancing gene editing specificity.
• Identified Cas9 hotspots and descriptors can guide future CRISPR-Cas9 system design.