3 results about "Protein surface" patented technology

This invention discloses nanoparticles based on transcytosis-activating ligands, their preparation method, and applications, particularly in the treatment of fundus diseases. The nanoparticles comprise an enzyme protein core and a polymer shell grown in situ on the protein surface. The polymer shell contains positively charged monomers, transcytosis-activating ligands, and neutral monomers. The nanoparticles can enhance the hydrolytic resistance of enzyme proteins (e.g., antioxidant enzymes), protecting them from degradation during transport and improving delivery stability. Simultaneously, the nanoparticles can also promote the efficient delivery of proteases (especially to the posterior segment of the eye), exhibiting good safety and showing excellent application prospects in the medical field.

The application discloses a protein reverse folding method and system fusing surface features and skeleton features, and relates to the technical field of protein structure prediction. The method comprises the following steps: obtaining protein main chain atomic coordinates and protein surface grid information; integrating skeleton geometry representation and surface geometry representation by using a joint representation learning mechanism, wherein a bipartite graph connecting the skeleton geometry representation and the surface geometry representation is constructed based on the spatial proximity of protein surface vertices and skeleton nodes in the Euclidean space, edge features in the bipartite graph are extracted based on the bidirectional interaction of skeleton and surface features, the edge features are updated hierarchically and iteratively, and a joint representation is obtained; and decoding the joint representation by using a multi-head attention-based decoder to generate an amino acid sequence of a target protein. The application can simultaneously consider the surface and skeleton structure features of the protein, and solves the problem of insufficient accuracy of existing reverse protein folding technology when processing surface exposed residues.

The present application belongs to the technical field of computational biology, and particularly relates to a protein surface multi-level modular representation method and an interaction interface prediction method. First, a protein manifold triangle mesh containing multi-modal physical and chemical characteristics is constructed. Then, based on the geodesic distance and local density field, the continuous surface is discretized into geometrically compact virtual residues using a restricted region growing algorithm. Next, an improved soft subspace clustering is used to mine atom patterns with specific functions, and a graph-guided manifold autoencoder is introduced to map them into high-dimensional embedding vectors. On this basis, a compatibility network is constructed by fusing statistical, semantic and physicochemical characteristics, frequent sub-patterns are mined, and the surface is resolved into a structured sub-pattern sequence. Finally, a graph neural network model is constructed for PPI prediction, and the prediction area is resolved into core modules and variable modules. The present application uses a unique four-level hierarchical structure and manifold geometry calculation to solve the problems of weak geometric perception and lack of interpretability in the prior art.