An Overview and Comparative Analysis of CRISPR-SpCas9 gRNA Activity Prediction Tools

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• CRISPR gene editing relies on efficient and specific guide RNA (gRNA) design.
• Numerous machine learning (ML) and deep learning (DL) tools exist for predicting gRNA activity, but a standardized evaluation is lacking.

Purpose of the Study:

• To systematically evaluate and compare the performance of ML and DL tools for predicting CRISPR-Cas9 editing efficiency.
• To assess the impact of different datasets and statistical metrics on model evaluation.
• To establish a centralized resource for CRISPR datasets and prediction tools.

Main Methods:

• Benchmarking seven ML and DL-based CRISPR-Cas9 editing efficiency prediction tools.
• Utilizing nine CRISPR datasets across six cell types and three species.
• Analyzing predictive performance using metrics like accuracy and Spearman correlation coefficient.

Main Results:

• Deep learning models CRISPRon and DeepHF demonstrated superior predictive accuracy and Spearman correlation.
• Performance varied across different datasets, highlighting the importance of data characteristics.
• A comprehensive resource, GuideNet, was compiled for sharing CRISPR datasets and in silico tools.

Conclusions:

• CRISPRon and DeepHF represent advanced DL models for predicting gRNA efficiency.
• Standardized evaluation across diverse datasets is essential for reliable CRISPR design tools.
• The GuideNet portal will facilitate data sharing and advance the development of more accurate CRISPR prediction models.