316 results about "Virtual screening" patented technology

The invention belongs to the field of bioinformatics, and relates to a molecular property prediction method based on multi-modal gating and comparative learning, which comprises the technologies of comparative learning, graph neural network, cross-modal alignment, gating attention and the like. Firstly, data standardization and graph construction are carried out, and molecular fingerprint embedding is extracted; secondly, a heterogeneous dual-channel graph coding architecture is adopted, one path captures atom short-range interaction through an attention mechanism, the other path integrates a molecular global structure and long-range dependence, and complementary molecular representation is generated; then, a cross-modal attention mechanism is introduced, bidirectional association of graph and fingerprint features is achieved, and modal weights are adaptively and dynamically distributed through a gating fusion module; and finally, a comparison pre-training strategy is adopted, a molecular graph and fingerprints are utilized to construct a sample pair, and discriminative molecular representation is learned on unlabeled data. According to the method, the accuracy of molecular property prediction is remarkably improved, and an efficient and reliable calculation tool is provided for virtual drug screening and lead compound optimization.

Disclosed herein is a method and system to generate 3D models of small molecules. The methods and systems are disclosed for an equivariant shape-based virtual screening operation that generates small molecule 3D structures that fit into and bind to a provided protein pocket. The model is configured to learn from existing 3D structures of small molecules in a data-driven manner using forward diffusion and backward diffusion processes to learn and represent the major characteristics of molecule 3D structures. Additional disclosure relates to the model that is further configured to learn from the 3D model of the protein pocket.

The invention provides an enzyme modification method based on artificial intelligence. The enzyme modification method comprises the following steps: training an AlphaFold2 model to obtain a glycosyltransferase tertiary structure prediction model; performing multi-sequence comparison on the protein and nucleic acid sequence database and the non-redundant protein feature sequence database based on the active point location recognition model to obtain a comparison result; initializing a matrix of which the length is equal to that of a comparison result, so as to construct a PSSM matrix and screen out potential mutation sites; according to the screened potential mutation sites, constructing an initial mutation library to form a virtual screening mutation library; performing structure reconstruction on the sequence in the virtual screening mutation library based on a glycosyl transferase tertiary structure prediction model so as to predict the glycosyl transferase tertiary structure and perform molecular docking and dynamic simulation on the glycosyl transferase tertiary structure; the method is used for analyzing binding pockets of proteins and ligands and determining dynamic interaction of amino acid residues and ligands in the binding pockets, and the contact number of each amino acid residue and the ligands is used for modifying the tertiary structure of the glycosyltransferase.

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.

A method for training a deep learning model configured for virtual screening of molecules for micro ribonucleic acid (miRNA) drug discovery includes creating a training dataset including a plurality of assay datasets. The method also includes training the deep learning model to learn a plurality of tasks; performing, using the deep learning model, inference for the respective task associated with at least one miRNA assay dataset; and evaluating respective performance metrics associated with the respective predictions for each of the plurality of tasks. The method further includes selecting a set of the plurality of assay datasets for training; and training the deep learning model to learn a set of the plurality of tasks using the set of the plurality of assay datasets. The trained deep learning model is configured to predict molecules capable of affecting a target miRNA.

The invention discloses a hexapeptide with a hypoglycemic effect as well as a preparation method and application thereof, and belongs to the technical field of biological medicines. The amino acid sequence of the hexapeptide is Ile-Trp-Asp-Pro-His-Phe, and the amino acid sequence of the hexapeptide is as shown in the specification. The novel hexapeptide IWDPHF with prominent DPP-IV inhibition ability is successfully identified from mulberry leaf proteolysis products through liquid chromatography-mass spectrometry and a virtual screening method, the hexapeptide IWDPHF shows the blood glucose reducing effect superior to that of part of known peptides in an in-vitro enzyme activity inhibition experiment and a zebra fish hyperglycemia model, and no obvious side effect exists; the compound has a good potential of being developed into a hypoglycemic functional product or a drug lead compound. The invention further provides the mulberry leaf peptide with the peptide fragment IWDPHF, and the mulberry leaf peptide can be applied to preparation of drugs or food for reducing blood sugar. The invention provides a new scheme and theoretical basis for treatment of diabetes, and has good market prospect and application potential.

The invention discloses a drug molecule virtual screening method based on combination mode analysis, and belongs to the field of artificial intelligence drug discovery. Using PLIP to calculate binding modes between the to-be-screened target and the small molecules, performing coding, using a DBSCAN algorithm to cluster the binding modes, and identifying compound populations of similar binding modes; in the screening stage, two key indexes are introduced: an energy index is formed according to a hydrogen bond compensation number and an AutoDock Vina score, the correlation between related residues and key site alpha carbon atoms in a binding mode is used as an activity index for evaluating the binding stability and characteristic activity between a target and a small molecule, and multi-target frontier analysis is performed on candidate molecules; according to the method, the efficiency and precision of virtual screening are remarkably improved, detailed analysis on intermolecular interaction is enhanced, the method is suitable for small-scale and large-scale molecular screening, the drug research and development cost is reduced, the drug research and development time is shortened, the diversity of virtual screening results is enriched, and potential candidate molecules are provided for subsequent experiments.

The invention discloses a virtual drug screening method and system based on adaptation during testing. The method comprises the following steps that a historical data set of protein pocket-small molecule pairing is collected; constructing a double-encoder framework consisting of a protein pocket encoder and a small molecule encoder; performing joint training on the double-encoder architecture based on the historical data set to obtain a small molecule screening model; and obtaining protein pocket data, and screening and sequencing the candidate small molecules corresponding to the protein pocket data based on the small molecule screening model to obtain corresponding active compounds. According to the method, the adaptation technology during testing is introduced into the virtual drug screening task for the first time, structural specificity dynamic adaptation of the model to an unknown sample is achieved on the premise of not depending on a test label, and the generalization performance is remarkably improved.

The invention discloses an endocrine disrupter quantitative activity prediction method, and relates to the field of endocrine disrupter virtual screening, and the method comprises the following steps: obtaining nuclear receptor in vitro experiment data, removing repeated data in the experiment data, and removing a compound set which does not contain simplified molecular linear input specification SMILES expression; a molecular fingerprint method is adopted to extract first-level, second-level and third-level feature structures of the compound. For each compound cluster, constructing a quantitative prediction model based on machine learning or quantitative read-stress for predicting the quantitative activity value of the compound; aiming at the low quantitative activity prediction efficiency of the endocrine disrupter in the prior art, multilevel structure characteristics of compounds are extracted, and corresponding quantitative prediction models are constructed for compound structure clusters of different scales. Through molecular docking and molecular dynamics simulation, the interaction mechanism and the like of the endocrine disrupter and the nuclear receptor are researched from the perspective of structural biology, and the efficiency is improved.

The invention is suitable for the technical field of virtual screening, and provides a molecular virtual screening method based on a parallel depth map Bayesian optimization framework. According to the method, a depth map Bayesian optimization framework is used for performing a virtual screening task; a parallel screening strategy is designed under a complex condition, that is, a parallel depth map Bayesian optimization framework is used for parallel operation on a plurality of computing nodes. In order to further reduce the time cost, the agent model is trained by using an incremental learning method, and meanwhile, the problem of disastrous forgetting is relieved by using an EMA strategy and an OCS sampling method. Compared with a serial depth map Bayesian optimization framework, the parallel depth map Bayesian optimization framework provided by the invention has the advantages that when the same screening task is carried out, the time cost is reduced by 75.1%, the training time of an agent model is reduced from 13.97 h to 4.49 h, and meanwhile, the screening performance is basically kept stable.

The invention discloses a composite double-protein hypoglycemic peptide as well as a preparation method and application thereof, and belongs to the technical field of animal and plant source double-protein active peptides. The method comprises the following steps: performing computer simulation enzyme digestion and virtual screening by utilizing bioinformatics to determine a protein raw material and protease, performing enzymolysis on animal and plant double proteins (soybean protein and casein) serving as raw materials by adopting neutral protease, and performing separation, purification and structural identification, thereby obtaining the protein. The protein peptide fragment with high alpha-glucosidase inhibitory activity is obtained by virtually screening molecular docking, and the amino acid sequences of the protein peptide fragment are shown as SEQ ID No: 6, SEQ ID No: 9 and SEQ ID No: 24. The compound double-protein peptide sequence provided by the invention has an inhibiting effect on the activity of alpha-glucosidase, can be used for preventing and treating diabetes mellitus, and can be used for long-term health care or treatment of people with impaired glucose tolerance or diabetics as a functional food ingredient.

The present disclosure provides a method, device, electronic device and storage medium for molecular screening, and relates to the fields of deep learning technology, graph theory technology and biological and information technology in the field of artificial intelligence technology. The method is as follows: obtain a first label map of the molecule to be screened and a second label map of the reference molecule, each molecule to be screened and the reference molecule form a molecule pair, for each molecule pair, obtain the mapping between the vertices of the first label map and the second label map and the conflict information between the mappings to generate a mapping map of the molecule pair, sample the mapping map, obtain the maximum weight fully connected subgraph of the molecule pair, and according to the maximum weight fully connected subgraph of each molecule pair, screen out the molecule to be screened with the greatest similarity to the reference molecule as the target molecule. A framework for molecular screening based on Gaussian boson sampling is proposed, so that Gaussian boson sampling can be used to efficiently implement the problem of ligand-based drug virtual screening and improve the efficiency of molecular screening.

The invention discloses umami peptide derived from litopenaeus vannamei and application of the umami peptide, and belongs to the technical field of biological active peptide. The number of the flavor peptides is seven, and the amino acid sequences of the flavor peptides are ASRM, RCNG, DNCY, GAEF, DEGF, EDMQF and PNRMPY respectively and are shown as SEQ ID NO.1-7 in sequence. The umami peptide can be used as or used for preparing a flavor enhancer. According to the present invention, the new umami peptide is explored from the litopenaeus vannamei meat protein sequence through the virtual screening, the molecular simulation and other technologies, such that the marine umami peptide library is enriched, the key amino acid combined with the umami peptide receptor and the action mechanism are researched, and the reference is provided for the research of the umami presentation mechanism of the peptides with different structures.

According to the screening method for the potential protein biomarkers and the drug targets of the gastric cancer, Mendel randomization analysis in a proteome range is adopted, the genetic causal relationship between circulating plasma protein and the risk of the gastric cancer is evaluated, and finally the remarkably related protein is identified. According to the screening method of the potential protein biomarker and the drug target of the gastric cancer, provided by the invention, the potential association between circulating plasma protein and the gastric cancer is systematically revealed by integrating Mendel randomization, single-cell RNA sequencing analysis, space transcriptome analysis, virtual drug screening, molecular docking, molecular dynamics simulation and other methods.

The invention discloses a preparation method of a dairy cow placenta extract CP1 and application of the dairy cow placenta extract CP1 as an antioxidant substance. The preparation method comprises the following steps: taking normal delivery healthy Chinese Holstein dairy cow placenta homogenate as a substrate, shearing the dairy cow placenta into pieces, homogenizing by adopting a tissue homogenizer, adding papain, and acting for 4.7 hours under the conditions that the pH is 6.5 and the temperature is 55 DEG C; the antioxidative peptide has the advantages that the antioxidative peptide with five potential types is screened by the aid of virtual screening processes after the antioxidative peptide is inactivated in boiling water bath for 10 minutes, affinity of the antioxidative peptide with Keap1 after molecular docking is higher than that of TX6, the affinity of ANNGKQWAEVF with Keap1 is the strongest, accordingly, the antioxidative peptide is marked as CP1, the CP1 is excellent in free radical scavenging capacity, lipid peroxidation inhibition capacity, transition metal chelating capacity and reducing power, and the antioxidative peptide with the antioxidative peptide can be used for preparing antioxidative peptide products. In addition, a Keap1 / Nrf2 pathway in oxidative stress cells can be activated, so that the oxidative stress of the cells is relieved, and the compound can be applied as an antioxidant substance.

The invention relates to the field of drug discovery and delivery collaboration, in particular to an autonomous evolutionary drug discovery and delivery collaboration method and system based on a large language model. The method comprises the following steps: aiming at a given biological target three-dimensional structure, generating a candidate molecular library which is complementary with a target pocket and gives consideration to druggability through an SE (3) isovariant hybrid generation model; according to a two-stage funnel type high-throughput virtual screening process, screening out the molecule with the highest comprehensive potential from the candidate molecule library; generating a customized delivery scheme through a three-stage process; in a digital twinborn model for simulating a real in-vivo tumor microenvironment, a targeted delivery process of a drug and carrier complex is simulated, and efficiency is evaluated; atomic-scale performance confirmation is carried out on a medicine, carrier and target point ternary system through molecular dynamics simulation. The invention is suitable for drug research and development.

The invention discloses a generative adversarial network optimization method for small molecule drug virtual screening, and belongs to the field of computer-aided drug design. The method comprises the following steps: constructing a small molecule drug data set; building a GAN basic model comprising a generator and a discriminator; performing multi-objective optimization training (optimizing chemical effectiveness, combining affinity and structural diversity) on the model through a joint loss function; generating candidate molecules by using the optimized model, and screening through a threshold value; and carrying out molecular docking verification on the screening result and outputting a final result. The quality of generated molecules is improved through multi-objective optimization, the drug research and development cycle is shortened, and the method is suitable for efficiently screening potential drug molecules.

The invention provides an application of liquiritin in preparation of a PPAR gamma receptor partial agonist. Through a structure-based high-throughput virtual screening technology, it is found that liquiritin can be used as a partial agonist of a PPAR gamma receptor, and the relative activation efficiency of liquiritin is 35.9% of that of rosiglitazone; in-vitro experiments prove that liquiritin can remarkably promote glucose uptake and consumption of HepG2 cells and does not induce adipocyte differentiation; in-vivo experiments prove that liquiritin can effectively improve the blood glucose level of an insulin resistance mouse model induced by high fat diet and reduce the level of serum proinflammatory factors. Compared with thiazolidinedione compounds, liquiritin not only can relieve and treat insulin resistance or related metabolic diseases, but also does not cause adverse reactions such as weight gain, fat accumulation and myocardial hypertrophy, provides a new candidate compound for developing safe and effective anti-diabetic drugs, and has a wide application prospect.

The invention relates to a virtual molecular screening method based on artificial intelligence and related equipment, and the method comprises the steps: obtaining SMILES of a plurality of to-be-screened molecules and series ligand molecules, and calculating molecular descriptors, parent nucleus structures and substituent groups of the series ligand molecules; generating a 3D conformation of the ligand molecule, performing molecular docking with the 3D conformation of the target protein, and calculating the protein-ligand interaction of the target protein; acquiring molecular activity experimental data of each ligand molecule; the molecular descriptor of the ligand molecule, the parent nucleus structure, the substituent group, the protein-ligand interaction and the corresponding molecular activity experiment data form training data to train an XGBoost model; the trained XGBoost model is incorporated into an SHAP algorithm for calculation and analysis, and SHAP values of the input data are obtained; making a molecular screening rule based on the SHAP value; and screening the molecules meeting the requirements from the plurality of to-be-screened molecules based on a molecule screening rule.

Virtual screening methods are disclosed for predicting interacting compounds, from a chemical library, for proteins. The methods using an Extremely Randomized Trees (Extra Trees) machine learning model in which parent node(s) represent a decision in relation to a target-compound complex, based on a plurality of features corresponding to a protein target of the complex and a plurality of features corresponding to a compound of the complex, and leaf nodes classify the target-compound complex as either interacting or non- interacting. The Extra Trees model outputs a probability that target-compound complex is interacting, based on aggregated classifications from each decision tree. Applying the Extra Trees model to a further protein target and compounds produces a probability that each compound will form an interacting complex with the further protein target. The compounds are then ranked based on the respective probabilities, and highest probability compounds are then explored using a docking model.

The invention provides a small molecule compound virtual screening method and system based on parallel Bayesian optimization, the method is applied to a plurality of computing nodes which work independently and have a communication relation, and the computing nodes comprise a master node and a plurality of slave nodes, the method comprises the following steps: S1, each computing node performs target property prediction on a target molecular library by using an agent model to obtain a prediction score and uncertainty, and calculates an EI score; s2, each computing node obtains small molecule compounds to be docked according to the descending order of the EI scores and the collection number, and molecular docking is carried out to obtain docking scores; s3, the main node determines the collection number of each computing node according to the docking score of each computing node; and S4, the main node updates the proxy model based on the to-be-docked small molecule compound of each computing node and the corresponding docking score, and the steps S1-S4 are repeated to perform multiple rounds of model training until a predetermined iteration round is reached. Through the method, idle resources can be avoided, and calculation time is saved.

The invention discloses a virtual screening method based on a Kv1.3 inhibitor, a drug lead compound and application, and relates to the field of pharmaceutical chemistry. The virtual screening method comprises the following steps: selection and pretreatment of a Kv1.3 protein crystal structure: obtaining a Kv1.3 protein-small molecule compound crystal structure from a Protein Data Bank database, and carrying out pretreatment on the Kv1.3 protein-small molecule compound crystal structure; determining a butt-joint active pocket; selecting a binding site; selecting and downloading a to-be-screened compound library; and virtual screening of the compounds: performing molecular docking by using virtual screening software to obtain the compounds of which the binding free energy (delta G) is negative and the absolute values are ranked from large to small. The compound Z239529546 has the beneficial effects that the compound Z239529546 screened by the method can reduce the expression of related inflammatory factors such as IL-6, IL-1beta, TNF-alpha and the like in cells and Kv1.3 protein in cell experiments, and has a certain effect of treating and / or preventing inflammation.

The invention discloses a key effect and a targeted intervention strategy of YTHDF1 in regulation and control of a hepatocellular carcinoma immunosuppression microenvironment. In the tumor-associated fibroblasts, YTHDF1 enhances the translation efficiency of OTUD5 mRNA through an m6A dependency mechanism, an up-regulated deubiquitination enzyme OTUD5 is directly combined with a tryptophan metabolic rate-limiting enzyme IDO1, ubiquitination degradation of the tryptophan metabolic rate-limiting enzyme IDO1 is inhibited, tryptophan is further driven to catabolism to generate kynurenine, regulatory T cell differentiation is promoted, and an immunosuppression microenvironment is formed. Spatial transcriptomics shows that YTHDF1 + CAFs and Tregs are significantly co-localized in tumor tissues and have metabolic interaction. Based on artificial intelligence virtual screening, it is found that a natural compound hesperidin can be specifically combined with a YTH structural domain of YTHDF1, an OTUD5-IDO1 signal axis mediated by the natural compound hesperidin is blocked, and the KYN level is remarkably reduced. Animal experiments prove that the tumor inhibition rate of combined treatment of hesperidin and anti-PD-1 is increased by 40% or above compared with that of a single drug group, and synergistic interaction is achieved by inhibiting Tregs infiltration and enhancing the activity of cytotoxic T lymphocytes.

The invention discloses a characterization optimization and soft label construction method for a drug-target interaction prediction model, and the method comprises the following steps: (1) constructing a molecular characterization space based on a Molformer pre-training model, carrying out the feature space similarity analysis of a negative sample through employing a K-Means clustering algorithm, and eliminating redundant samples through setting a dynamic threshold value, thereby achieving the equalization of positive and negative samples; (2) designing a soft label optimization mechanism based on a logarithmic function, converting a pKd value in a DAVIS data set into soft labels continuously distributed in a (0, 1) interval, and reserving binding strength gradient information to enhance the sensitivity of the model to a weak binding compound; and (3) integrating the above data enhancement strategies, and training a drug target interaction prediction model through a five-fold cross validation method to improve the model prediction performance. The method verifies the significant advantages of the method in relieving the problems of data offset and information loss, and also provides a high-robustness data processing framework for virtual drug screening of deep learning.

The invention discloses a virtual screening method of a tyrosinase inhibitor and a composition with a whitening function. The method comprises the following steps: firstly, constructing and preprocessing a data set of a tyrosinase inhibitor, then respectively establishing a QSAR classification model and a QSAR regression model, and carrying out model training and parameter optimization by utilizing various machine learning algorithms. Virtual screening is performed on a large-scale compound library through the two models, candidate compounds with high inhibitory activity on tyrosinase are preliminarily screened out, and the activity of the candidate compounds is further verified through an in-vitro tyrosinase inhibition test. Experimental results show that the screened compound has a remarkable inhibition effect on tyrosinase and has potential application prospects in whitening cosmetics and freckle-removing medical products. The method is low in cost and short in time consumption, the novel tyrosinase inhibitor can be efficiently found, and a powerful tool is provided for research and development of raw materials with whitening effects.

This invention relates to systems and methods for adaptive discovery and mixed-variable optimization of synthesizable microelectronic materials, and applications of the same. Specifically, an exemplary system includes a virtual screening (VS) module to extract information from literatures of a knowledge base by text mining, a ML-assisted conceptual exploration (CE) module to identify candidate material families for the specific class of compound materials based on the extracted information via a combination of ML models and to generate exogenous models of objective functions f(x, y) and constraint functions g(x, y), and an adaptive discovery (AD) engine to generate and optimize design of the newly discovered compound materials. The AD engine includes a mixed-variable ML module, a mixed-integer optimization (MIO) module, and a high-fidelity evaluation (HFE) module, which are iteratively and sequentially executed.

The application provides a drug virtual screening method and device based on deep learning, wherein the method comprises the following steps: inputting all candidate compound small molecules in a candidate molecule database into a pre-trained molecular encoder respectively to obtain molecular vectorization representation of the candidate compound small molecules; constructing an index structure corresponding to the candidate compound small molecules based on the molecular vectorization representation of the candidate compound small molecules; inputting a protein target to be matched into a pre-trained protein target encoder to obtain protein target vectorization representation corresponding to the protein target; and matching the protein target vectorization representation based on the index structure to obtain a target compound small molecule corresponding to the protein target. The method establishes a full-amount mapping function of the target and the molecule through a vector calculation mode, realizes high-throughput virtual screening in a second-level calculation time, realizes rapid virtual screening of a full-amount candidate molecule library, improves the precision of the virtual screening, and increases the possibility of drug discovery.

The invention discloses a mulberry leaf antihypertensive active peptide as well as a preparation method and application thereof. According to the method, mulberry leaves are used as raw materials to prepare mulberry leaf albumin, enzymolysis parameters of enzymolysis time, pH, enzyme-substrate ratio and substrate concentration are optimized through a single factor experiment and an orthogonal experiment, and optimal enzymolysis conditions for preparing the ACE inhibitory peptide crude product are provided; the high-activity ACE inhibitory peptide is identified through further purification, polypeptide sequence determination and virtual screening, the ACE inhibitory peptide has extremely high solubility and good ACE inhibitory activity, and the IC50 values of the two peptides VPSCFDLTGK and RLPDFHGL with the highest activity are 8.23 [mu] mol / L (8.765 [mu] g / mL) and 23.01 [mu] mol / L (23.58 [mu] g / mL) respectively. The active peptide can be used as an angiotensin converting enzyme inhibitory peptide, and achieves the purpose of reducing blood pressure by inhibiting ACE activity.

The invention discloses application of glycoside compounds in prevention and treatment of plant pathogenic viruses, and belongs to the technical field of pesticide antivirus. According to the invention, targeting tobacco mosaic virus capsid protein is taken as an example, and small molecule picroside I (A0) and ZINC85593196 (B0) containing sugar rings are screened through virtual screening, molecular dynamics simulation, molecular mechanics and MM / GBSA free energy calculation. On the basis of further optimization, six compounds of 4-methylumbelliferone-beta-D-xyloside (A1), geniposide (A2), 4-methylumbelliferone acyl-beta-D-glucopyranoside (A3), 4-methylumbelliferone acyl-beta-D-glucopyranoside (A4), catalpol (A5) and aesculin (sesquihydrate) (A6) which are similar to the A0 in structure are obtained, and in-vivo and in-vitro biological activity evaluation is carried out. The active component screened by the method has a good inhibition effect on plant viruses.

Embodiments of the present disclosure provide a molecular property prediction method, related device and medium. The method splits and recombines a first molecule in an unlabeled molecule dataset to obtain a new second molecule, determines a training sample based on the first molecule and the second molecule, and optimizes a molecular encoder using the training sample. Then, the molecular property prediction model based on the optimized molecular encoder is trained using a labeled molecule dataset to achieve accurate prediction of the molecular property. Embodiments of the present disclosure can fully utilize the substructure information inside the molecule to improve the accuracy and generalization ability of the prediction, optimize the molecular representation ability of the molecular encoder, and enable the target molecular encoder to better understand the combination relationship between the molecular fragments, thereby improving the representation quality and prediction accuracy of complex molecular structures. Embodiments of the present disclosure can be applied to drug discovery, material science, molecular virtual screening and other scenarios.