多模态跨视野下的蛋白质翻译后修饰位点预测方法和系统
By constructing a multimodal cross-field feature representation and a cyclic iterative fusion method, the problems of insufficient information integration and long-range dependence in the existing technology are solved, and more efficient prediction of post-translational modification sites is achieved, improving prediction accuracy and stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SUZHOU UNIV
- Filing Date
- 2026-04-30
- Publication Date
- 2026-07-17
AI Technical Summary
Existing methods for predicting post-translational modification sites suffer from insufficient information integration, inadequate long-range dependency capture, and insufficient robustness in utilizing multimodal information, making them difficult to adapt to complex real-world biological data scenarios.
We employ a multimodal, cross-field approach, constructing feature representations based on sequence and structural modalities, combining them with RIDCGA for iterative fusion, and using a cross-field large-kernel attention module and stacked autoencoders to extract complementary features of proteins, thereby achieving efficient modeling of long-distance amino acid dependencies.
It improves the accuracy and stability of post-translational modification site prediction, enhances the prediction accuracy for imbalanced and incompletely labeled data, and improves the model's generalization ability in complex biological data scenarios.
Smart Images
Figure CN122157757B_ABST