36 results about "Molecular graph" patented technology

A foundation model for performing molecular-level tasks by learning multimodal data in the form of one-dimensional text and two-dimensional graphs, and a system therefor, according to an embodiment of the present invention, enable various molecular-unit tasks such as chemical reaction prediction, molecular attribute prediction, and natural language description generation to be effectively processed through a single foundation model. In addition, it is possible to increase prediction accuracy of the model by maximizing utilization of two-dimensional molecular graph information, and automatically generate, on the basis of statistical sparsity, high-quality descriptive text that emphasizes core and distinctive features of each molecule.

A formulation graph convolution network (F-GCN) with multiple GCNs assembled in parallel and connected to filters and an external learning architecture is able to predict the effectiveness of a formulation. Input into the multiple GCNs are molecular structures of formulants, which are processed as molecular graphs and output as molecular descriptors. The molecular descriptors are filtered by normalized ratios or fractions of the ingredient molecules in a formulation, such as a battery electrolyte or solvent. A formulation descriptor combines the filtered molecular descriptors to arrive at a predicted performance for the formulation, such as the battery capacity for an electrolyte formulation, by an external learning architecture. F-GCN may use a pre-trained GCN with physico-chemical properties of known molecular structures.

The application discloses a compound water solubility prediction method based on enhanced molecular graph representation learning, and aims to improve the prediction accuracy and robustness. The technical scheme is: a compound water solubility prediction system based on enhanced molecular graph representation and double-branch graph neural network fusion is constructed, which is composed of a data preprocessing and enhanced molecular graph construction module and a water solubility prediction module (containing a double-branch fusion graph neural network model). The data set required for constructing the training prediction system is constructed. The double-branch fusion graph neural network model contains two branches of GraphSAGE and GIN, and the model is trained by using the AdamW optimization method combined with the learning rate scheduling strategy, gradient clipping and early stopping mechanism, and the trained prediction system is obtained by loading the model parameters with the best performance on the validation set. The trained prediction system is used to predict the compounds to be predicted, and the water solubility prediction value of the compound is obtained. The prediction accuracy of the application is significantly improved compared with the traditional method.

The application relates to the technical field of energetic material design and performance prediction, and relates to an energetic material multitask property prediction method. Standardized data sets are constructed by collecting energetic material sample data and performing molecular graph data enhancement processing on a training subset; molecular SMILES strings are respectively parsed into molecular topological graphs and word sequences, three-dimensional graph neural network is used to encode atomic space position information to extract molecular local topological features, a Transformer encoder is used to extract molecular global long-range correlation features, and a dynamic weighting mechanism is used to realize adaptive fusion of the two types of features, so that the depth and comprehensiveness of feature mining are improved; a multitask prediction model with a shared encoding layer, multiple independent decoding layers and a task decoupling constraint mechanism is built, an end-to-end training is completed by using a multitask loss function, the problems that existing multitask methods are difficult to balance the internal conflicts between energy performance and safety performance and are difficult to simultaneously predict multiple key indicators are solved, and the simultaneous prediction of multiple performance indicators is realized.

The application discloses a traditional Chinese medicine small molecule Gibbs free energy prediction method and system based on deep learning. The application receives a traditional Chinese medicine small molecule in the form of a SMILES string or a molecular structure file, and after structure verification and standardization processing, the traditional Chinese medicine small molecule is converted into a molecular graph representation. The molecular graph is input into a deep learning model. The model captures the molecular structure characteristics and interaction relationship through a graph representation learning mechanism, refines the representation through multi-layer updating, and finally outputs the predicted Gibbs free energy value of the traditional Chinese medicine small molecule through a pooling operation. The test set determination coefficient R2 is greater than or equal to 0.99. The application improves the calculation efficiency and prediction accuracy while maintaining accuracy. In addition, the model of the application supports general equipment deployment, and has strong deployment flexibility.

The application provides a kind of molecular graph-oriented element level representation learning method and system, based on graph diffusion model, original molecular graph is added with reconstruction, edge reconstruction error is calculated, and original molecular graph is divided into multiple element subgraphs according to edge reconstruction error;Atom view and element view are constructed, and graph neural network encoder is used to encode atom view and element view respectively, to obtain atom embedding matrix and element embedding matrix;Based on atom view and element view, positive and negative sample pairs are constructed, and barlow twin double view contrast loss function is used to optimize atom embedding matrix and element embedding matrix, to realize cross-granularity representation alignment.The application extracts molecular element through diffusion model self-supervision, and solves the systematic defects of traditional method in cross-granularity structure modeling, semantic guided contrast learning and multi-level representation fusion through atom-element double view alignment, significantly improves the semantic completeness and task generalization ability of molecular representation.

The present application relates to a kind of molecular structure perception chemical reaction kinetics equation symbol regression method, for solving the problem that existing symbol regression technique ignores species chemical information, causes poor modeling effectiveness.The method adopts encoder-decoder architecture, and the encoding part is provided with double branch: one branch is handled concentration time series by Transformer, another branch utilizes graph neural network to encode molecular graph, and structural characteristics are extracted.After bimodal feature fusion, it is directly returned to the kinetics equation of ordinary differential equation form by the Transformer decoder with cross attention mechanism.Training data is generated by synthesis: based on the data set of elementary reaction, SMILES is converted into molecular graph, and the kinetics equation is deduced according to mass action law, and then concentration trajectory is generated by random sampling and numerical integration, and training sample comprising molecular graph, trajectory and equation is formed.The present application significantly improves the accuracy, interpretability and data utilization efficiency of symbol regression in reaction kinetics modeling.

The application discloses a kind of metal organic compound energy prediction methods based on deep learning, belong to compound property prediction technical field.It includes the following steps: obtaining the molecular structure information of metal organic compound and calculating electronic structure characteristics, constructs molecular graph and encodes generation geometry feature vector;Global fusion features are output through multimodal feature fusion model;Total energy value and molecular structure are predicted through multi-task prediction network;The mean and variance of the total energy value are obtained by statistical prediction, and the energy prediction value and confidence interval are obtained.The application constructs molecular graph and extracts geometric features, and simultaneously utilizes cross attention and gate fusion network to realize the deep interactive fusion of geometric and electronic structure features, solving the problem of inaccurate feature representation of metal organic compounds and poor multimodal feature fusion effect of traditional methods.

The present application belongs to the technical field of chemical structure recognition, and particularly relates to a chemical structure recognition method and a recognition system. The method comprises the following steps: obtaining an original data set containing chemical structure images based on historical literature, and generating an image segmentation data set and an image recognition data set according to the original data set; converting PDF format literature into a plurality of to-be-recognized images for literature needing chemical structure recognition, recognizing chemical structures in the plurality of to-be-recognized images, and extracting chemical structure images; generating an image recognition learning data set and an image recognition test data set respectively according to the image recognition data set, recognizing chemical atoms in the chemical structure images through an image recognition model, and recognizing hyper texts; reasoning and constructing a chemical molecule graph based on the chemical atoms and the hyper texts, and analyzing and outputting chemical structure formulas conforming to SMILES or InChI specifications. The present application can extract machine-readable chemical structure formulas from PDF format documents.

The application discloses an unsupervised domain adaptation method based on molecular graph structure alignment, adopts a graph neural network with a double-branch structure to extract implicit local semantics and explicit functional substructure semantics of a molecular graph, extracts and aggregates Motif substructures based on a predefined functional group template, obtains explicit features representing high-order structure patterns of the molecular graph, introduces an adversarial domain alignment module, narrows the distribution difference between a source domain and a target domain in graph structure representation through a domain discriminator, proposes a pseudo-label screening mechanism with collaborative consistency, screens high-confidence pseudo-label samples from the target domain by using the consistency of double-branch prediction results to participate in training, performs class distribution alignment, designs a cross-branch structure contrast learning module to constrain the structure semantic representations extracted by the double branches to be consistent, and constructs a joint loss function and performs optimization.

The application relates to the field of drug discovery and molecular design, and particularly relates to a molecular attribute prediction method and device based on a heterogeneous graph neural network and a medium. The method obtains structure information of a to-be-predicted molecule and a target attribute category to be predicted; extracts atomic-level features to construct atomic nodes and atom-atom edges; connects corresponding atomic nodes and pharmacophore nodes; connects virtual molecule nodes and attribute nodes, so as to construct a task-specific heterogeneous molecular graph containing atomic nodes, globally shared pharmacophore nodes, attribute nodes, virtual molecule nodes, atom-atom edges, atom-pharmacophore edges, molecule-attribute edges and pharmacophore-attribute edges; inputs the graph into a pre-trained heterogeneous graph neural network model; generates a global representation vector through a cross-layer adaptive attention mechanism, and outputs a prediction result of the target attribute category. The method can capture the complex hierarchical relationship in a molecule and realize effective molecular attribute prediction under a small sample condition.

This invention discloses a method and apparatus for drug molecule generation based on a latent space multi-constraint diffusion model, relating to the field of smart medical technology. The method includes: S1, obtaining the two-dimensional structure of a known molecule and representing it as an initial feature matrix and adjacency matrix; S2, learning the two-dimensional structural features of the known molecule, obtaining the final feature matrix, and converting it into a latent space; S3, constructing and optimizing the latent space multi-constraint diffusion model; S4, generating a new hidden space; S5, analyzing the new hidden space to generate a new atom list and a new adjacency matrix, obtaining a new molecule; S6, analyzing whether the new molecule is a valid molecule and outputting the generation result. This method solves the problem that current graph neural network models cannot capture the unique topological properties of molecular graphs well; and avoids the problem of poor data adaptability that occurs when performing a latent space multi-constraint diffusion model directly on a discrete graph structure, enabling the generated molecule to achieve a balance between various attributes while satisfying each attribute constraint.

The application discloses a photovoltaic material structure performance relationship analysis method based on a graph neural network, and belongs to the technical field of the cross of artificial intelligence and new energy materials. The method comprises the following steps: acquiring degradation process data of a photovoltaic material under the coupling action of multiple working conditions, wherein the degradation process data comprises a multi-scale structure snapshot sequence and corresponding working condition parameter information; obtaining structure feature representation results corresponding to each structure snapshot, fusing the structure feature representation results with the corresponding working condition parameter information, and obtaining a working condition self-adaptive molecular graph sequence; extracting multi-scale structure features representing the degradation behavior of the photovoltaic material, and forming a corresponding time sequence feature sequence; obtaining time sequence state features representing the degradation evolution process of the photovoltaic material; constructing a photovoltaic material structure performance relationship model; obtaining corresponding performance attenuation results; and identifying key defect structures, key degradation stages and dominant degradation mechanisms affecting performance attenuation in the degradation process of the photovoltaic material.

The application discloses a method for predicting ADMET properties of drug compound molecules based on deep learning fusion of molecular graphs and molecular point cloud, and comprises the following steps: obtaining SMILES code of a candidate drug molecule for which drug molecular property prediction is needed; processing the SMILES code of the candidate drug molecule; generating a corresponding two-dimensional molecular graph based on the number of atoms and the number of chemical bonds in the SMILES code of the candidate drug molecule, extracting atom node features and edge features of atomic bonds in the drug molecule from the SMILES code of the candidate drug molecule, adding the atom node features and the edge features of the atomic bonds in the drug molecule into the two-dimensional molecular graph, and obtaining a drug molecular graph; extracting corresponding point cloud features of the candidate drug molecule based on the SMILES code of the candidate drug molecule; and inputting the drug molecular graph and the corresponding point cloud features of the candidate drug molecule into a trained deep learning model to obtain an ADMET property prediction result of the candidate drug molecule.

This invention relates to the field of enzyme function prediction technology, specifically to an enzyme function prediction method based on deep contrastive learning. The method includes: acquiring amino acid sequence data of the target enzyme and molecular structure data of the reaction substrate; performing sequence embedding encoding on the amino acid sequence data to generate sequence feature vectors, and performing graph structure modeling on the reaction substrate molecular structure data to generate molecular graph feature vectors; fusing the two data across modalities to generate a joint feature representation containing spatial complementarity features of enzyme-substrate binding sites and electron transfer features of catalytic active centers; calling a pre-trained contrastive learning encoder to perform contrast sample mining on the joint feature representation, generating a set of positive and negative sample pairs; constructing a contrastive loss function based on this set, and updating the encoder parameters through backpropagation to obtain an optimized model; inputting the data into the optimized model, and outputting the predicted catalytic function category of the target enzyme. This method improves the accuracy and stability of enzyme function prediction.

This invention discloses an estrogen receptor target identification and toxicity prediction system and its operation method based on graph neural networks and molecular simulation technology. The system converts compound structures into molecular graphs, constructs atomic node features and chemical bond edge features / types, inputs these into a graph neural network for characterization learning, and outputs the binding probabilities of estrogen receptor subtypes such as ESR1, ESR2, and GPER1, as well as predicted toxicity endpoints such as logAC50 / AC50. Based on the predicted target, the system automatically selects the receptor structure and adaptively generates docking box parameters. It then calls AutoDock Vina to complete molecular docking to obtain binding energies and optimal conformations. The system also displays the three-dimensional complex, interaction statistics, public database annotation information, and toxicological explanations based on a large language model in an interactive interface. This system achieves a high-throughput, interpretable, integrated "prediction-docking-annotation" analysis workflow for ER targets.

The application belongs to the technical field of bioinformatics and artificial intelligence, and specifically discloses a bad drug interaction prediction method and system based on double-view motif learning, which comprises the following steps: constructing atom-level molecular graphs and motif-level molecular graphs of drug pairs, extracting atom-level features and aggregating them into motif-level molecular graphs; constructing a similarity-guided double-view interaction module to learn the interaction relationship between motifs within a drug and the interaction relationship between motifs across drugs to obtain motif representations after double-view interaction, and then constructing a global interaction vector and calculating attention feature vectors of two drugs; and inputting the attention feature vectors after splicing into a multilayer perceptron to output a prediction result of drug interaction. The application solves the problems of the existing DDI prediction method, such as lack of explicit modeling of substructures with chemical significance, lack of guidance of chemical prior knowledge for learning of cross-drug substructure interaction, and inherent information loss and ambiguity of uniqueness in motif-based modeling.

This invention belongs to the field of bioinformatics and relates to drug-target interaction (DTI) identification technology, specifically providing a drug-target prediction method based on structure-aware message passing. This invention proposes a structure feature extraction module based on sequence generation trees to model the multi-scale structure of drugs and extract features, incorporating structural information into the representation update process of the drug molecular graph, thereby enriching the molecular graph representation. Furthermore, multiple protein correlation matrices are used to fuse information such as protein structure and physicochemical properties, enhancing protein representation. Finally, an interaction feature extraction module based on transformers and message passing neural networks is used to better capture the bidirectional interaction between drugs and targets, thereby improving the accuracy of drug-target prediction.

This invention discloses a method for predicting the function of bioactive peptides based on multi-view, multimodal representation learning. The method includes: extracting amino acid sequence information of peptides using multi-scale dilated convolutional CNN and bidirectional LSTM; extracting structural and functional features of peptide molecules using an ESM-2 model; processing molecular fingerprint information using convolutional CNN and Mamba structures; extracting topological information of the peptide molecular graph using traditional convolutional CNN, and processing node features using graph convolutional neural networks. All these multi-view features are ultimately fused into an aggregated feature representation, which is then passed through a fully connected layer and a sigmoid function is applied for multi-label classification. By concatenating and fusing the extracted features, a comprehensive peptide molecule feature representation is formed to predict various bioactive properties of the peptide. This multi-view, multimodal feature integration method not only enhances the model's predictive ability but also improves its flexibility and accuracy when processing complex biological data.

The application belongs to the technical field of liquid crystal material and data driving material design, and discloses a liquid crystal monomer performance prediction method and a mixed crystal performance prediction method based on GAT, which represents the liquid crystal monomer as a molecular graph, encodes the chemical bond connection relationship and chemical environment by using a graph attention network, and introduces the polarization rate, the dipole moment and the orbital energy level and other quantitative calculation physical information in the pre-training stage; a PIML molecular prediction model is established under a structure perception data division strategy, and accurate prediction of key properties such as birefringence, dielectric parameters and clearing point is realized. On the basis of constructing a mixed crystal prediction model based on KernelSVR prediction, forward mixed crystal performance prediction or reverse mixed crystal formula generation is realized, which significantly reduces the formula search cost and improves the research and development efficiency.

The application belongs to the field of material design calculation method, and discloses a kind of anion exchange membrane molecular design method based on chemical knowledge enhanced double-channel graph attention network.The method embeds chemical priori knowledge into molecular graph representation, and learns hydrophilic ion segment and hydrophobic non-ion segment features through double-channel graph attention network respectively to explicitly represent hydrophilic and hydrophobic microphase separation;Meanwhile, introduce feature linear modulation to dynamically integrate experimental conditions into molecular features, realize multi-scale fusion of molecular structure information and external conditions.The method can realize the level prediction and evaluation of OH⁻ conductivity and alkali stability of anion exchange membrane, and through attention weight visualization, analyze the structure-activity relationship between key structural units and performance, provide interpretable basis for molecular structure optimization and new structure construction, so as to establish the iterative process of "prediction-analysis-structure modification / new structure construction-re-prediction", realize molecular design closed loop.

The invention discloses an optical chemical structure identification method based on a discrete diffusion model, and relates to the field of artificial intelligence and chemical informatics cross technology, and the method comprises the following steps: S100, constructing an optical chemical structure identification network; s200, training an optical chemical structure recognition network; s300, visual feature extraction is carried out; s400, constructing an initial sequence and dividing the initial sequence into continuous sequence blocks; s500, iterative denoising and dynamic truncation are carried out; and S600, reconstructing the molecular image and performing standardized output. According to the method, efficient and accurate identification of the optical chemical structure is realized, the defects in the prior art are overcome, and the identification accuracy and the reasoning efficiency are improved.

The application belongs to the technical field of multi-agent, and particularly relates to a glycomics agent construction method and system based on three-level coding and a storage medium, which maps sugar chain sequences, molecular graph topologies and three-dimensional space conformations to the same vector space through a three-level coding mapping network to generate multi-modal semantic representation vectors; inputs the representation vectors and multi-source spectral graph data into a multi-modal agent to output predicted sugar chain structures with structural unit confidence through cross-modal fusion and uncertainty quantification; generates research hypotheses automatically based on the prediction results and the confidence on a glycomics knowledge graph through k-hop random walk; generates recommended experimental parameters in an experimental parameter space through Bayesian optimization; issues the recommended experimental parameters to an automated experimental platform and receives results, evaluates deviation through KL divergence calculation, automatically updates knowledge graph weights and fine-tunes the multi-modal agent; and the application breaks the whole unmanned link from original spectral graphs to new knowledge discovery.

The invention discloses a genome scale metabolic network model deletion reaction prediction and filling method and system, and the method comprises the steps: obtaining a plurality of genome scale metabolic network models from a metabolic network database, and constructing a data set with a supervision label; extracting and fusing the molecular sequence and molecular map characteristics of the metabolite to obtain an initial characteristic vector; constructing a metabolic directed graph and a metabolic reaction hypergraph based on a positive reaction sample, performing feature directivity enhancement by using the directed graph, and extracting high-order topological information through a hypergraph convolutional neural network to obtain final feature representation of metabolites; on the basis of the feature representation, adopting an attention mechanism to predict candidate reaction confidence and training a model; and finally, screening a high-confidence reaction from the candidate reaction pool and filling the target model with the high-confidence reaction. According to the method, the metabolite multi-dimensional molecular characteristics and the network high-order topological information are deeply fused, the prediction accuracy is remarkably improved, the method does not depend on experimental data, and the method is suitable for efficient metabolic network model correction and optimization.

The application discloses a general drug property prediction method and device based on a conditional perception graph neural network, and the method comprises the following steps: constructing a condition-specific training data set; constructing a prediction model, wherein the prediction model is a double-branch architecture, and comprises a semantic extraction module, a double-branch molecular graph encoder and a conditional mixed expert prediction head; the double-branch molecular graph encoder comprises a conditional perception branch and a structure reservation branch; a phased pre-training strategy is adopted to train the prediction model based on the training data set, thereby obtaining a trained general molecular property prediction model, wherein the phased pre-training strategy comprises physicochemical property perception pre-training and large-scale biological activity pre-training; and the trained general molecular property prediction model is used to process a molecular chemical structure in combination with task and experimental condition description, and a property prediction value of the molecule under specific task and experimental conditions is output. The application can effectively integrate truncated information.

The application relates to the cross field of artificial intelligence and material chemistry, and discloses a special polymer reverse design method and system, which solves the problems of low reverse design efficiency of polymers with specific physical and chemical properties and dependence of existing polymer research and development on trial and error experience. The scheme receives target physical and chemical property requirements; constructs a reinforcement learning intelligent agent to splice polymer fragments with preset chemical groups as an action space; uses a molecular graph neural network fusing a global context vector to extract graph structure features and predict macro properties; combines the difference between the predicted properties and the requirements with chemical synthesis feasibility scoring to construct a multi-objective reward function, iteratively updates a policy network, and automatically generates a new polymer monomer combination and topological structure meeting the requirements. The application is used for intelligent research and development of special polymer materials and shortens the new material discovery cycle.

This invention relates to the field of healthcare information technology, and more particularly to a method and system for predicting adverse drug reactions based on graph neural networks. The method includes: receiving and verifying the molecular structure information of candidate drugs to obtain verified candidate drug molecular structure information; using a two-layer molecular graph encoding mechanism to perform feature encoding on the verified candidate drug molecular structure information to obtain a molecular representation vector of the candidate drug; inputting the molecular representation vector of the candidate drug into a multi-label adverse drug reaction prediction network to obtain the original logits vector for each adverse drug reaction category; calibrating the confidence level of the original logits vector for each adverse drug reaction category using a temperature scaling method to obtain the calibration probability and corresponding confidence level for each adverse drug reaction category, and outputting an adverse drug reaction prediction report. The technical solution of this invention enables reliable confidence quantification of the prediction results, improving the accuracy of adverse drug reaction prediction.