A CRISPR / Cas9 off-target prediction method based on a multi-scale convolutional neural network
By constructing a multi-scale convolutional neural network model and the Dice loss function, the imbalance problem of the CRISPR/Cas9 off-target dataset was solved, the accuracy of off-target prediction was improved, and more reliable support was provided for gene editing and gene therapy.
CN116312770BActive Publication Date: 2026-07-21SHANTOU UNIV
View PDF 0 Cites 0 Cited by
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANTOU UNIV
- Filing Date
- 2023-03-03
- Publication Date
- 2026-07-21
AI Technical Summary
Technical Problem
The existing CRISPR/Cas9 off-target dataset suffers from data imbalance, making it difficult to effectively predict off-target scenarios.
Method used
A multi-scale convolutional neural network model is constructed, combining one-hot encoding and the Dice loss function. The dataset is processed by merging and undersampling, and the model is trained to improve prediction accuracy.
Benefits of technology
It improves the accuracy of CRISPR/Cas9 off-target prediction, provides a more accurate and robust algorithmic support, and facilitates the realization of precision genome editing and gene therapy.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116312770B_ABST
Abstract
The embodiment of the application discloses a CRISPR / Cas9 off-target prediction method based on a multi-scale convolutional neural network, comprising the following steps: constructing a CRISPR / Cas9 off-target benchmark data set, carrying out one-hot encoding on the guide RNA sequence and the DNA sequence of the benchmark data set respectively, and obtaining two binary matrices. Then, the two matrices are spliced up and down to be combined into one binary matrix; a CnnCRISPR model is constructed based on a multi-scale convolutional neural network and combined with a Dice loss function of a class imbalance; the basic network parameters of the CnnCRISPR model are trained by using the benchmark data set; for an independent test set, the guide RNA sequence and the DNA sequence pair of the data set are encoded, input into the trained CnnCRISPR model for prediction, and whether the guide RNA sequence and the DNA sequence interaction produces off-target is analyzed. By using the application, the advantages of the multi-scale convolutional neural network in extracting sequence features of multiple scales and the Dice loss function in solving data imbalance are combined, and the application has the advantages of high prediction accuracy and strong robustness.
Need to check novelty before this filing date? Find Prior Art