Phosphocholine cytidylyltransferase mining method based on deep learning model and application thereof
By combining deep learning models with structural prediction and rational design, the problem of insufficient catalytic activity and expression efficiency of existing CCT enzymes in the synthesis of citicoline was solved. This enabled the efficient identification and modification of highly active, low-homology CCT enzymes, thereby improving the efficiency of citicoline biosynthesis.
Patent Information
- Application Number
- CN202610249550.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-03
- Publication Date
- 2026-07-24
- Estimated Expiration
- 2046-03-03
AI Technical Summary
Existing CCT enzymes lack sufficient catalytic activity, thermal stability, and heterologous host expression efficiency in the industrial production of citicoline. Traditional enzyme mining methods struggle to identify high-activity, low-homology enzymes from massive metagenomic data, and the screening results are highly homogeneous.
Using a deep learning model-based approach that combines protein language models and structure prediction models, we screened for highly catalytically active and soluble CCT candidate enzymes through Seq-to-Seq search, Hidden Markov Model alignment, transmembrane segment and signal peptide prediction, structure prediction, and ligand molecule docking. We then obtained mutants with improved performance through rational design.
This method efficiently identifies highly active CCT enzymes with low homology within a sequence space of hundreds of millions of sequences, overcoming the limitations of traditional methods, significantly improving the screening success rate, obtaining novel CCT enzymes and mutants, overcoming the problem of poor expression adaptability, and improving the efficiency of cytidine diphosphate choline biosynthesis.