496 results about "Protein structure" patented technology

The invention discloses a drug target activation and inhibition relation prediction method based on a depth map neural network, and aims to improve the modeling precision and prediction performance of an activation or inhibition action mechanism between a drug and a target. According to the method, on the basis of a fine-grained graph interaction modeling mechanism, multi-scale structural characteristics of drug molecules and three-dimensional space structural information of protein residue levels are fused, and a heterogeneous interaction graph between drugs and proteins is constructed. The method comprises the following steps: firstly, acquiring a drug-target sample with an activation / inhibition tag through a public database, predicting a protein structure by utilizing AlphaFold2, and constructing a protein residue map and a drug molecular map; multi-scale structure semantic representation is obtained through sub-graph decomposition, atomic-scale feature extraction and graph neural network coding of drug graph features; protein graph node features are combined with context embedding generated by a pre-training language model, DSSP coding, secondary structure spectrum and atomic structure features are constructed, and edge features are designed based on the geometrical relationship between residues. Then, based on constraints such as spatial distance and biochemical similarity, a fine-grained mapping relation between drug atoms and protein residues is established, an interaction graph is constructed, and coding is carried out through a GraphSAGE network; and finally, fusing the interacted multi-source embedding, and completing the prediction of the activation / suppression relationship through a multi-layer perceptron. A cross entropy loss function, an Adam optimizer and hyper-parameter grid search are adopted in model training; in the evaluation stage, five-fold cross validation and an independent test set are adopted, and indexes such as the accuracy rate, the recall rate, the F1 score, the specificity and the Morse correlation coefficient are used for comprehensively evaluating the performance of the model. Experimental results show that compared with an existing method, the method has the advantages that the prediction accuracy and mechanism interpretability are remarkably improved, and the method has good generalization ability and application prospects and is suitable for multiple fields of drug action mechanism research, new drug discovery and the like.

The invention discloses a DNA binding residue prediction method based on a multi-modal protein language model in the technical field of bioinformatics. The DNA binding residue prediction method based on the multi-modal protein language model comprises the following steps: sequentially using tools to respectively obtain feature files of a protein sequence P to be subjected to DNA binding residue prediction; inputting all the feature files into the ESM3 to obtain an embedded matrix M; the method comprises the following steps: processing a protein sequence into a residue sample, constructing an isovariant graph neural network model according to a protein structure diagram, constructing a data set by utilizing the protein sequence of known DNA binding residues, and training the constructed network; and inputting a residue sample of the protein sequence into the trained model to obtain a DNA binding residue of the protein sequence. According to the DNA binding residue prediction method based on the multi-modal protein language model, the efficiency and accuracy of DNA binding residue prediction are improved.

The invention provides a drug target prediction method based on a 3D structure and a multilevel attention mechanism, and belongs to the technical field of drug target prediction. The technical problem to be solved is to provide a drug target prediction method based on a 3D structure and a multi-level attention mechanism. According to the technical scheme, the method comprises the following steps: collecting data information of an interaction tag among a related drug, a target sequence and a drug target protein, and processing and extracting SMILES character string characteristics of the drug; learning the numeralized protein amino acid sequence by using a convolutional neural network, completing feature coding of a target spot on the sequence through multi-layer convolution and pooling operation, and extracting target spot sequence feature data; a PDB file is read and modeled into a pocket graph, feature calculation is carried out on the pocket graph, a final protein structure representation code is generated, and structural feature data of target protein is extracted by modeling the PDB file; the method is applied to drug target prediction.

The invention discloses a small molecule ligand drug screening method and system based on affinity prediction, and belongs to the technical field of biological medicine. The invention aims to solve the technical problem of low drug screening precision caused by molecular expression limitation, geometric invariance deficiency and insufficient multi-modal information fusion when virtual drug screening is carried out by using protein-ligand affinity. Comprising the following steps: acquiring ligand and protein structure information, and preprocessing to obtain coordinates and a feature matrix of ligand / pocket / residue; performing comprehensive representation, multi-feature flow self-adaption, geometric algebraic multi-layer perception and feature alignment processing on the feature matrix to obtain corresponding feature space representation; performing cross attention fusion and multi-scale interactive learning processing on the feature space representation in sequence to obtain fusion features; inputting the fused features into a multi-scale interactive learning module, and outputting final features; and finally, predicting the binding affinity of the ligand and the protein according to the fusion characteristics to obtain a binding affinity value.

The invention discloses cytosine deaminase as well as a related biological material and application thereof, and belongs to the technical field of protein. The technical problem to be solved by the invention is to excavate natural cytidine deaminase without sequence preference and improve the base editing efficiency and range. The invention provides a 15th-143rd site protein of which the amino acid sequence is SEQ ID NO.2. The invention also provides a preparation method of the protein. The invention also provides application of the protein and related biological materials thereof in single base editing. According to the application, metagenome big data is utilized, comprehensive and systematic information mining analysis is carried out on novel cytidine deaminase, and a protein L85 which is highly similar to a known DddA protein structure but low in sequence similarity is selected from the novel cytidine deaminase for functional verification. The protein L85 has high catalytic activity as cytidine deaminase, is not limited by target sequence preference, can be widely applied to gene editing in vitro and in cells / tissues, and has good application potential and development value.

The invention belongs to the field of drug design, and particularly discloses a virtual screening method of a small-molecule inhibitor targeting YTHDF1 protein, which comprises the following steps: acquiring a YTHDF1 protein structure, and carrying out homologous modeling to generate various conformation proteins; small molecules combined with the YTHDF1 protein are obtained, bait molecules are generated, and ligand molecules are formed; a ligand molecule is in butt joint with a binding site on the YTHDF1, and the compound is characterized through PLEC fingerprints; the method comprises the following steps: constructing a specificity scoring function by using a machine learning algorithm, carrying out scoring screening on a compound to be screened and a compound butted with YTHDF1 molecules, carrying out affinity prediction, and screening out the small-molecule inhibitor targeting YTHDF1. According to the virtual screening method provided by the invention, the YTHDF1 targeting molecules with high affinity can be accurately screened on a large scale, candidate molecules are provided for research and development of medicines for treating YTHDF1 related diseases, and a foundation is laid for clinical tests and medicine optimization.

The invention relates to the field of protein-polypeptide interaction prediction in bioinformatics, in particular to a protein-polypeptide binding site prediction method based on graph attention and multi-head attention. The method mainly comprises the following steps: (1) collecting protein and polypeptide compound PDB structure information from an RCSB PDB database, and extracting sequence information of the protein and polypeptide compound PDB structure information; (2) extracting protein sequence information by using IUPred2, ProtBERT and ESM-2 (Extensible Subscriber for Mobile Communications); a ProtBERT method and an Integer method are used for extracting polypeptide sequence information; the method comprises the following steps: extracting protein structure information through biopython; (3) establishing a GAT model with residual connection to analyze a protein structure, and extracting features between amino acid nodes; (4) constructing a Circulate Block module, and carrying out deep extraction and fusion on protein and polypeptide information through four layers of Mti-head Attention and Dual Attention; and (5) finally, through a Final Attention, a linear layer and a Softmax layer, mapping the features to two dimensions to represent the interaction probability of each residue site.

The invention provides a method set system for predicting dynamic structure change of protein A beta 42 based on a deep learning method. The method comprises the following steps: extracting PDB of each frame from a molecular dynamics track; the 2DIR of each PDB is calculated; splitting the PDB according to the functional characteristic region; an alpha-carbon distance matrix distance map of each PDB is calculated, and an alpha-carbon distance matrix distance map of each PDB is calculated; according to the trained deep learning model, predicting an alpha-carbon distance matrix distance map of each region; constructing a protein three-dimensional structure; merging fragments and comparing the fragments with a three-dimensional structure: aligning the optimized three PDB segments, and merging the three PDB segments into a complete PDB protein three-dimensional structure; and finally, solving the RMSD error between the predicted PDB and the real PDB. According to the method, a division and treatment strategy is adopted, three sections of different regions in a protein structure are predicted by utilizing 2DIR, and finally, a three-dimensional structure is successfully constructed. According to the method, a new method is provided for exploring the structure and dynamic change of the Abeta42, a new view angle for researching the Alzheimer's disease is opened up, and important information is also provided for drug design.

The invention relates to the field of molecular targeted therapy, and discloses an application of an SLC16A5 (MCT6) small-molecule inhibitor MCT6-Ai7-2 in preparation of a medicine for treating acute myeloid leukemia (AML). The inhibitor takes an SLC16A5 protein structure predicted by Alphafold as a target spot, and is obtained through compound database screening, molecular docking and druggability optimization. An in-vitro experiment proves that MCT6-Ai7-2 can remarkably inhibit proliferation of AML cell lines such as U937 and MOLM-13, induce cell apoptosis and retard a cell cycle, has an inhibiting effect on a primary AML patient specimen and is relatively low in toxicity to normal cells; in-vivo experiments show that the compound is effective and has good safety in AML model mice. In addition, the MCT6-Ai7-2 and the vinca can be combined to synergistically enhance the inhibition effect on vinca drug-resistant cells, and by reducing the expression of anti-apoptotic protein MCL-1, the activation of pro-apoptotic factors BIM and tBID is promoted to play a role. The invention provides a novel targeting drug and strategy for treatment of AML (especially drug-resistant or recurrent patients).

The invention belongs to the field of construction of a protein structure on a cryoelectron microscope density map, and relates to a method for constructing a protein structure from a cryoelectron microscope density map by combining de novo modeling and structure prediction, a computer device, a readable storage medium and a program product. According to the method, the atomic probability and the amino acid type are predicted through the three deep neural networks respectively, full-atom optimization is carried out, the output result is combined with the graph theory, optimization and geometric algorithms to jointly assist protein structure modeling, and protein structure information can be mined to the maximum extent. According to the method, de novo modeling can be carried out in the absence of full-length structure data of the protein monomer, integrated modeling can also be carried out under the condition of inputting the full-length structure data of the protein monomer, the application scene is wide, the structure of the protein compound can be automatically constructed in a cryoelectron microscope density map with middle and high resolution, and the method is suitable for popularization and application. And for a region with poor local resolution in a traditional method, the modeling precision can be remarkably improved.

A method for functional antimicrobial peptide design is provided. The method includes running pretrained protein large language models as a generator and enhancing a sample candidate. The enhancing a sample candidate includes performing an automatic pipeline based on machine learning methods and a plurality of bioinformatics methods and is configured to perform protein inverse folding and computational protein sequence designing. The computational protein sequence designing includes running a pretrained deep learning-based protein structure model, an autoregressive pretrained protein sequence model, and a deep learning-based alignment model. The pretrained deep learning-based protein structure model is configured to learn three-dimensional structures of the proteins and output latent structure embeddings in a high dimensional space. The autoregressive pretrained protein sequence model includes a protein language model to automatically generate protein sequences and is trained to predict next amino acid in a protein sequence based on a preceding sequence(s) of amino acids.

The invention relates to a protein implicit binding site prediction method based on deep learning. The method comprises four stages of data acquisition, feature engineering, model training and verification application. The method comprises the following steps: firstly, acquiring initial data from a protein database, performing data screening and sequence redundancy elimination, performing standardization processing on a PDB structure file, and generating a sample; then, constructing a composite feature space containing basic protein features and evolutionary conservative features; thirdly, feature dimension reduction is achieved through a hierarchical feature conversion and integrated feature selection method, and a model is constructed; finally, verification and application show that the method obtains 99.44% prediction accuracy on 954 non-redundant protein structures, ROC-AUC and PR-AUC both reach 0.9998, and the overlapping rate with a BioLiP database is 44.4%. The method can accurately identify potential drug binding sites, especially shows good generalization ability for drug targets difficult to prepare, and has important application prospects.

The application discloses a method for recovering active glycoprotein Patatin from potato starch processing waste juice, and belongs to the technical field of protein extraction. The method is characterized in that the active Patatin protein is efficiently separated by using the combination of ammonium sulfate fractionation precipitation, anion exchange chromatography, Con A affinity chromatography and gel chromatography, and the obtained Patatin protein retains the original protein structure and biological function activity. Compared with the prior art, the method has better separation purity and higher efficiency, and the obtained Patatin protein has complete protein higher structure and function activity, and has a better application scene.

The invention relates to the technical field of atlas generation, in particular to a children cross-allergen risk dynamic atlas generation system, which comprises the following steps: firstly, determining sensitivity indexes according to immune response conditions of each type of allergens in all allergy detection processes; according to the similarity of each allergen and other allergens in two dimensions of allergen protein structure and IgE level, combining the sensitivity index, and comprehensively representing the cross reaction risk index of each allergen in each allergy detection process; then comprehensively determining the cross-allergy risk degree of various allergens in each season period according to the increasing trend of the cross-allergy risk index curve in each season period and the overall size of the cross-allergy risk index in combination with the characteristic that the cross-allergy reaction possibly caused by the season change is dynamically changed; and the cross allergen risk dynamic map generated in real time according to the cross allergen risk degree is more accurate.

The invention relates to the technical field of biology, and particularly discloses a recombinant protein for improving the curative effect of an ADC drug and application of the recombinant protein. The recombinant protein comprises a tumor cell surface targeting structure, a cell transmembrane structure and toxin molecules, the toxin molecule is connected and fused with the tumor cell surface targeting structure and / or the cell transmembrane structure; the tumor cell surface targeting structure is a protein structure capable of being specifically combined with a tumor cell surface target spot; and the cell penetrating structure is a protein structure for mediating the recombinant protein to penetrate through a cell membrane to enter the cell. The tumor cell surface targeting structure is used for specifically recognizing a target cell surface target spot, the killing activity of a conventional antibody is brought into play, then toxin molecules are brought into tumor cells through the cell transmembrane structure, toxin is released, the tumor killing effect is achieved, the ADC drug curative effect is improved, and the ADC drug application prospect is wide. The transmembrane efficiency of the cell transmembrane structure can be improved to 50-90%, so that the traditional ADC drug treatment window is improved.

In one aspect, a method is provided for identifying a residue in a particular conformation of a protein molecule that is capable of forming a stable bond between a compound and the protein molecule. The method can include identifying one or more residues from the conformation of the protein molecule to include in a perturbation set, the one or more residues exhibiting a protection factor that satisfies one or more thresholds. One or more residues may be excluded from the perturbation set based at least on the spatial arrangement and / or orientation of the one or more residues. One or more rounds of molecular dynamics simulations may be performed, the results of which are indicative of interactions between the compound and the residues in the perturbation set. The residues in the perturbation set may be identified as forming a stable binding between the compound and the protein molecule based at least on the results of the molecular dynamics simulation.

The invention provides a vaccine target screening system based on calculation model simulation. The vaccine target screening system comprises a multi-source heterogeneous database, wherein the multi-source heterogeneous database integrates and standardizes pathogenic genes, protein structures, literatures and experimental data; the feature calculation module calls a calculation biological model to carry out structural analysis, immunogenicity simulation and stability prediction; the intelligent screening and sorting module applies a multi-objective optimization algorithm to perform parallel evaluation and outputs optimal target spots; a structure iteration optimizer automatically iteratively corrects the optimized target spots to generate a high-potential variant library; and the process suitability simulation module couples the variants with the preparation formula and the process parameters to simulate production storage behaviors and feeds back an optimization target. According to the invention, efficient screening and optimization of vaccine targets can be realized, the accuracy and efficiency of target screening are improved, the research and development cost is reduced, and the research and development process of vaccines is accelerated.

The invention provides a system for designing and optimizing D-type biomolecules and a method and application thereof. The system comprises a stereoisomerization conversion module, a conjugate skeleton generation module, a sequence design module and an optimization module. The method comprises the following steps: firstly, carrying out stereoisomerization conversion on a target protein structure to generate a stereoisomer structure, and generating a constrained conjugate by utilizing a modeling algorithm under the structural constraint of the stereoisomer structure; on the basis of the generated conjugate skeleton, predicting an adaptive amino acid sequence, and then performing stereoisomerization conversion to obtain the corresponding D-type biomolecule conjugate. And meanwhile, an evolutionary Bayesian optimization algorithm is adopted to carry out iterative optimization on the biomolecule sequence. According to the method, structural constraint and an advanced optimization algorithm are fully utilized, efficient design and optimization of the D-type biomolecules are achieved, high universality and expandability are achieved, and a powerful tool is provided for drug development and biotechnology research.

The invention relates to the technical field of gene identification, in particular to a method for identifying a horizontal transfer gene based on a model error, which comprises the following steps of: acquiring a class group protein data set from a public database, and performing data cleaning on the class group protein data set to obtain a class group protein data set; the class group protein data set comprises a plurality of protein sequences of different class groups; performing structure prediction on each protein in the cleaned class group protein data set, and extracting a multi-dimensional feature vector based on a prediction result; a classification prediction model is constructed, the multi-dimensional feature vectors are adopted for training, and the output of the classification prediction model is the class group label probability; and predicting a protein structure in a to-be-detected protein sequence, extracting a multi-dimensional feature vector with the same dimension as the training stage based on a prediction result, inputting the multi-dimensional feature vector into the trained classification prediction model, and identifying a potential HGT gene according to a prediction error of the classification prediction model.

The invention aims to provide a method and equipment for constructing a protein sequence prediction model, and the method comprises the steps: determining surface point cloud information and sample protein sequence information corresponding to corresponding sample protein based on sample protein information, the sample protein information comprises protein sequence information and protein structure information, and sequence information belonging to the surface of a sample protein in the sample protein sequence information is masked; based on the surface point cloud information corresponding to the sample protein and the sample protein sequence information, a corresponding protein sequence prediction model is obtained through training, and the protein sequence prediction model comprises a structure encoder, a sequence encoder and a corresponding decoder. According to the method, the information of the protein surface is introduced in the construction of the protein sequence prediction model for sequence prediction, so that the model prediction accuracy can be effectively improved, and the surface precision of the predicted protein sequence is improved.

A method for operating a protein structure-based candidate discovery system may include: displaying a first screen for receiving input of a plurality of ligands to be docked to a target protein; calculating predicted binding energy by sequentially performing molecular docking on the plurality of ligands based on settings input by a user via the first screen; displaying a first list on a second screen, the first list containing rows including a user interface control for user selection, names of the plurality of ligands, and the calculated binding energy values; predicting binding affinity for the rows selected by the user via the user interface control from the first list displayed on the second screen; and displaying a second list on a third screen, the second list containing rows including the names of the plurality of ligands, values of the binding energy, and values of the predicted binding affinity.

The invention relates to the technical field of aquatic product preservation, and discloses a natural eutectic solvent for maintaining the quality of frozen penaeus vannamei boone and application, the natural eutectic solvent is composed of choline chloride, sorbitol and water, the molar ratio of choline chloride to sorbitol is (1-5): (1-5), and the mass of water is 15-25% of the total mass of choline chloride and sorbitol. The invention also provides an application of the solvent, and the application comprises the following steps: soaking the penaeus vannamei meat in the solvent, and then freezing and storing the penaeus vannamei meat. A supramolecular network system is constructed through hydrogen-bond interaction between choline chloride and sorbitol, growth of ice crystals in the freezing storage process is inhibited, a protein structure is stabilized, and loss of unfrozen juice is reduced, so that the quality of the penaeus vannamei, such as texture, color and flavor, is maintained. The natural deep-eutectic solvent is natural and safe in component, low in heat, simple to prepare and convenient to apply, the application limitation of a traditional anti-freezing agent is solved, and the natural deep-eutectic solvent has a good industrial prospect.

A varicella-zoster virus (VZV) nanoparticle protein, and a preparation method therefor and a use thereof. The VZV protein comprises a partial or full sequence of an amino acid sequence of an extracellular region of a VZV gE glycoprotein, with W at position 200 mutation to C and L at position 245 mutation to C in the amino acid sequence of the extracellular region of the VZV gE glycoprotein. By means of the rational optimization design of the amino acid sequence of the VZV gE protein by means of protein genetic engineering, the VZV gE recombinant protein modified with amino acid mutations has increased stability and immunogenicity compared with the VZV gE protein. Moreover, by further designing the protein structure, the VZV gE protein is repeatedly displayed on the surface of ferritin nanoparticles with the desired epitopes exposed, thereby further enhancing immunogenicity.

The invention discloses a polypeptide activity screening method and system based on artificial intelligence and an entity library. The method comprises the following steps: acquiring physicochemical properties and structural information of polypeptide molecules, physicochemical properties and structural information of target proteins and experimental data of combination of the target proteins and the polypeptide molecules from the entity library; processing experimental data of combination of the target protein and the polypeptide molecules; performing training verification on the initial model based on the physicochemical property and structure information of the polypeptide molecules, the physicochemical property and structure information of the target proteins and the processed experimental data to obtain a prediction model; and inputting the to-be-screened protein structure and the polypeptide sequence into the prediction model to carry out affinity prediction so as to complete screening of polypeptide activity. Through the scheme of the invention, a mode of combining artificial intelligence and entity library real big data is realized, the artificial intelligence model is trained by utilizing a large amount of real experimental data, the magnitude of compound screening is expanded through the intelligent model, the screening cost is rapidly reduced, and the experimental efficiency is improved.

The invention discloses a tyrosinase inhibitor, a preparation method thereof and application of the tyrosinase inhibitor in preparation of a product for inhibiting tyrosinase activity. Based on the protein structure of the TYR, the TYR inhibitor which is novel in structure and higher in curative effect and selectivity is found by adopting reasonable drug design and structural modification strategies.

The invention discloses a fusion protein, an artificial antigen presenting molecule and application of the artificial antigen presenting molecule in NKT amplification. The fusion protein comprises CD1d and beta2M linked by a linker. According to the fusion protein, CD1d and beta2M are connected through a linker, the fusion protein can be used for specific amplification of NKT cells by artificial antigen presenting molecules, and the potential safety hazard that gene mutation may be caused by irradiation of PBMC is avoided. Moreover, the amplification effect of an artificial antigen presenting molecule adopting the fusion protein is better than that of a dimer artificial antigen presenting molecule formed by spontaneous polymerization of CD1d and beta2M, meanwhile, the fusion protein is used for the artificial antigen presenting molecule, and specific amplification of NKT cells can be realized by adopting a very short scaffold protein structure; the production cost of the artificial antigen presenting molecule is reduced while the specific amplification effect of the NKT cells is improved.

The invention discloses a green preparation method of hilsa herring roe-derived peptide, the hilsa herring roe-derived peptide and application of the hilsa herring roe-derived peptide, the hilsa herring roe-derived peptide is prepared by the following steps: cleaning and dehydrating hilsa herring roes with 10% saline water to reduce the water content of the roes, then freeze-drying and crushing to prepare freeze-dried hilsa herring roe powder, and degreasing the protein of the hilsa herring roes by a supercritical CO2 extractor to obtain the hilsa herring roe-derived peptide. Rancidity of the hilsa herring roes caused by lipid oxidation is effectively prevented; a hilsa herring egg solution is subjected to ultrasonic treatment, the protein structure of the hilsa herring eggs is uniformly damaged, and specific protease is adopted for enzymolysis, so that the enzymolysis time can be effectively shortened, the enzymolysis efficiency can be improved, and the hilsa herring egg source peptide with antioxidant activity and ACE inhibitory activity is obtained. The anti-aging test of the hilsa herring roe-derived peptide in a nematode animal model shows that the hilsa herring roe-derived bioactive peptide can obviously prolong the life of nematodes, and the average life of the nematodes can be prolonged by 33.33%. The method disclosed by the invention is of great significance in fully utilizing shad roe resources, developing functional foods and anti-aging products and realizing high-valued utilization of characteristic aquatic resources.

The invention relates to the technical field of molecular docking, in particular to an automatic molecular docking and screening analysis method which comprises the following steps: S1, aligning protein structures; s2, processing the receptor protein file; s3, carrying out batch molecular docking; s4, calculating the distance between the substrate and the catalytic site; s5, performing combined screening on the combination energy and the spatial relationship; and S6, summarizing and visualizing results. According to the method, a large-scale molecular docking task can be efficiently and accurately completed, and powerful technical support is provided for the fields of drug design, enzymology research and the like.

The invention relates to a cell line capable of stably expressing an orthogonal aminoacyl tRNA synthetase / tRNA pair, constructs an HEK293T cell line capable of stably expressing an EcTyr orthogonal aminoacyl tRNA synthetase / BsTyr tRNA pair and discloses a construction method of the cell line, and the construction method comprises the following steps: stably integrating the orthogonal EcTyr aminoacyl tRNA synthetase and BsTyr tRNA gene into an HEK293T cell genome by virtue of a PiggyBac transposon system and stable plasmid transfection; according to the constructed cell line genome, the EcTyr orthogonal aminoacyl tRNA synthetase and the BsTyr tRNA gene are integrated, so that a stable platform is provided for efficiently introducing non-natural amino acid at any site of protein. The method has the effect of further laying an experimental foundation for exploration of novel protein structures and functions and research and development of protein drugs with certain specific functions.

The invention relates to a protein solubility prediction method and system based on GCN and an improved attention network. The method comprises the following steps: constructing a protein sequence data set; the method comprises the following steps: acquiring a protein three-dimensional structure file through AlphaFold3, and constructing a node feature vector; extracting position embedding, bidirectional projection direction features, a cross-node atom pair distance and a rotation relation, and constructing an edge feature vector; extracting global physicochemical property feature vectors of the protein; and constructing protein graph structure data, inputting the data to the graph convolutional network module and the improved graph attention network module, and performing training to obtain a graph convolutional network and an improved attention network model to perform protein solubility prediction. By extracting node and edge features and protein global physicochemical property features, subsequent model training is facilitated; and the two-way projection and the rotation relation enable the model to accurately sense the spatial orientation between residues, so that the model has more quantitative understanding on the protein folding and gathering driving force, and the accuracy of protein solubility prediction is improved.