Development of a Plant-Specific CRISPR Off-Target Prediction Framework: Benchmarking, Algorithm Optimization, and Regulatory Applications

CRISPR-Cas genome editing has revolutionized plant biotechnology, yet accurate prediction of off-target effects remains challenging due to the predominant focus of existing computational tools on mammalian systems. This dissertation addressed the critical need for plant-specific off-target prediction frameworks through a comprehensive mixed-methods computational research design. The study pursued four primary objectives: (a) developing a curated database of experimentally validated plant off-target events, (b) characterizing off-target patterns across plant species, (c) identifying plant-specific features influencing off-target activity, and (d) establishing design requirements for an optimized prediction framework. Through systematic literature mining and database integration, a comprehensive plant off-target database was constructed comprising 92 experimentally validated off-target records spanning 32 plant species and 19 unique guide RNAs, achieving 96.2% data completeness. Analysis revealed that SpCas9 dominated plant applications (93.5%), with crop species representing 85.9% of records. A strong negative correlation between mismatch number and off-target frequency was observed (r = -0.760, p < 0.001), with the seed region (positions 1-12) demonstrating significantly lower mismatch tolerance than the non-seed region. Epigenomic data integration encompassed 29 datasets with 1,802 samples across 11 species, revealing substantial variation in chromatin accessibility patterns. Plant genome sizes ranged from 135 Mb to 17,000 Mb, necessitating species-specific computational approaches. The findings established evidence-based requirements for framework development, including hybrid machine learning architectures, plant-specific feature engineering, and regulatory-compliant validation protocols. This research contributes the first comprehensive plant-specific off-target resource, provides empirical foundations for algorithm optimization, and offers practical guidelines for regulatory applications in plant genome editing.