CRISPR-MFH: A Lightweight Hybrid Deep Learning Framework with Multi-Feature Encoding for Improved CRISPR-Cas9...

Area of Science:

• Biotechnology
• Genomics
• Bioinformatics

Background:

• CRISPR-Cas9 gene editing offers significant potential but faces challenges with off-target effects.
• Current deep learning models for off-target prediction often underutilize sequence pair information and can be overly complex.
• Increasing model parameter size leads to greater complexity, hindering practical application.

Purpose of the Study:

• To develop a novel, efficient, and accurate deep learning framework for predicting CRISPR-Cas9 off-target effects.
• To address the limitations of existing models in leveraging sequence pair information and managing complexity.

Main Methods:

• A novel multi-feature independent encoding method to represent gRNA-DNA sequence pairs as three distinct feature matrices, minimizing information loss.
• Development of CRISPR-MFH, a lightweight hybrid deep learning framework integrating multi-scale separable convolutions and hybrid attention mechanisms.

Main Results:

• The proposed encoding method effectively captures critical sequence features.
• CRISPR-MFH demonstrates superior or comparable performance to state-of-the-art models across multiple benchmark datasets.
• The framework achieves high accuracy with significantly fewer parameters, indicating improved efficiency.

Conclusions:

• This research presents a novel approach to enhance deep learning for CRISPR-Cas9 off-target detection.
• The developed method and framework offer a more practical and effective solution for managing gene-editing risks.