Comparative Performance of Recent AI Models for CRISPR Prediction

ОБЪЯСНИМЫЕ И БЕЗОПАСНЫЕ ПЛАТФОРМЫ ИСКУССТВЕННОГО ИНТЕЛЛЕКТА ДЛЯ ГЕНОМИКИ CRISPR: ДОСТИЖЕНИЯ, ОГРАНИЧЕНИЯ, СИСТЕМЫ МОДЕЛИРОВАНИЯ И ТРАНСЛЯЦИОННАЯ ИНТЕГРАЦИЯ НА ОСНОВЕ БЛОКЧЕЙНА

Model	Architecture	Dataset	Task	AUC	F1-Score	Key Limitation
DeepCRISPR	CNN + Autoencoder	sgRNA datasets from HEK293T, HCT116, HeLa, HL60	On-target prediction	~0.80	0.76	Limited interpretabilit y
TransCRISPR	Transformer + CNN	CRISOT	gRNA efficiency prediction	~0.84	0.81	Computationa l complexity
DeepMEns	CNN + LSTM + Attention	CRISPRclean SE	Off-target prediction	~0.86	0.83	Limited crossspecies validation
CRISTA	Random Forest	GUIDE-seq datasets	Off-target prediction	~0.79	0.74	Limited contextual genomics
CRISTA-IT	Hybrid ML	Integrated genomic datasets	Off-target identification	~0.82	0.78	Limited scalability