Multi-network protein complex detection method and system based on training at test time
By constructing a multi-network data model and a reinforcement learning agent, the problem of failing to effectively utilize known complex information in existing technologies is solved, thereby improving the accuracy of protein complex detection and enabling the fusion of information from multiple biological networks to enhance detection precision.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ANHUI UNIV
- Filing Date
- 2026-02-13
- Publication Date
- 2026-06-05
AI Technical Summary
Existing technologies fail to effectively utilize known complex information in protein complex detection, and relying solely on protein interaction networks cannot integrate other biological information such as gene co-expression networks, resulting in insufficient detection accuracy.
A test-time training-based multi-network protein complex detection method is adopted. By constructing a multi-network data model, introducing a reinforcement learning agent, and using the protein network topology consistency and cross-network mapping consistency loss functions, the model is trained in combination with a known set of complexes, and the model parameters are dynamically adjusted to integrate information from multiple biological networks.
It improves the accuracy of protein complex detection, enabling real-time fusion of information from multiple biological networks to precisely locate target complexes and enhance detection accuracy.
Smart Images
Figure CN122157804A_ABST