65 results about "Molecular Fingerprint" patented technology

The invention discloses a drug resistance prediction method and system based on comparative learning and multi-modal fusion, and the method comprises the steps: firstly generating a molecular map and a molecular fingerprint based on the SMILES of a target drug, and extracting the molecular features of the drug through a comparative learning model constructed through combining a map attention network and a map convolution network; and then, acquiring protein expression, gene expression and metabolic expression data from the target tissue cells, extracting modal features through a deep convolutional network, a Transform encoder and a multi-dimensional attention network, and realizing adaptive fusion of the multi-modal features through a heterogeneous interactive attention mechanism. And finally, jointly inputting the fused multi-modal features and drug molecular features into a multi-layer sensor to realize high-precision prediction of the drug resistance of cells to drugs. By introducing a contrast learning and multi-modal feature fusion mechanism, the characterization capability and prediction precision of the model are effectively improved, and efficient and reliable support can be provided for drug screening and clinical decision making.

The invention relates to a molecular property prediction method and system based on multi-task pre-training and multi-modal fusion. The method comprises the following steps: collecting a data set and performing molecular conversion; performing multi-task pre-training, which comprises the following steps: generating a heterogeneous enhanced view; constructing a pseudo label; performing comparative learning according to the structure enhanced view and the heterogeneous enhanced view, and constructing a maximum similar task; capturing semantic differences among molecules according to the pseudo labels; carrying out multi-modal fusion, namely introducing functional group structure information, and extracting molecular sequence characteristics based on Transform and Mamba2 to obtain fused molecular multi-modal representation; and analyzing and predicting the classification or regression task. A heterogeneous enhanced view is established in a multi-task pre-training stage, multi-task self-supervision is performed in combination with multi-granularity features of molecular fingerprints, and functional group structure information is introduced in a multi-modal fusion stage, so that deep cross-modal interaction is realized, downstream prediction performance is improved, and candidate drug screening and molecular property evaluation processes are accelerated.

The invention relates to the technical field of artificial intelligence drug research and development, in particular to a molecular ADMET property prediction algorithm based on multi-mode and multi-scale characteristics. According to the algorithm, multi-scale feature extraction is carried out on a small molecule SMILES character string by constructing a multi-modal feature fusion framework, and training and prediction are completed on 73 ADMET properties covering absorption, distribution, metabolism, excretion, toxicity and general characteristics. The invention provides a multi-modal and multi-scale feature fusion strategy, atomic features, molecular fingerprint features and physicochemical property features are integrated, and a multi-scale feature set covering molecules is formed. A progressive information extraction architecture is combined with a graph neural network (GCN), a gated loop unit (GRU) and an attention mechanism to extract features, so that the accuracy of ADMET property prediction is remarkably improved. In a word, the method marks an important step for predicting the ADMET property of the medicine.

The invention provides a new pollutant classification prediction method and a related device, and relates to the technical field of new pollutant classification. Comprising the following steps: collecting chemical information (molecular structure, molecular fingerprint and physicochemical property) of a target pollutant and carrying out standardization treatment; preliminary classification is calculated based on Tanimoto similarity; simulating a migration path and a conversion behavior of the pollutants in the environment by using a graph neural network; multi-mechanism comprehensive evaluation is carried out by comprehensively considering durability, mobility and biological accumulation; introducing an environmental factor correction model, and correcting scores according to parameters such as temperature, pH value and dissolved oxygen concentration; integrating the molecular structure, the migration path, the mechanism characteristics and the environment correction result, and calculating a final classification prediction value; and dividing risk levels (high, medium and low risks) according to predicted values. Pollutant risks are comprehensively assessed through a machine learning technology, the assessment precision and reliability are improved, and a scientific basis is provided for pollution prevention and control.

The invention provides a cell-targeted drug preliminary screening model training method, a drug preliminary screening method and equipment, and the method comprises the steps: obtaining molecular fingerprint feature data corresponding to each compound sample with label information to form a corresponding original data set, and carrying out the data screening of the original data set in a distance measurement mode, and training a regression model based on a graph neural network to learn the structural topology of each compound sample, and training the model as a cell-targeted drug preliminary screening model for predicting whether the compound has potential cell-targeted drug activity or not. According to the application, the comprehensiveness of molecular characterization can be effectively improved in the training process of the cell-targeted drug preliminary screening model, and the generalization ability of the trained cell-targeted drug preliminary screening model can be effectively improved, so that the application universality of the trained cell-targeted drug preliminary screening model can be effectively improved; and the accuracy and the reliability of a potential cell targeted drug activity prediction result predicted by the model can be improved.

The invention relates to the technical field of mine water disaster prevention and control, in particular to a mine water disaster treatment slurry stone body recognition system and method based on an infrared spectrum. According to the technical scheme, the mine water disaster treatment slurry stone body recognition system and method based on the infrared spectrum comprises a rock debris preprocessing and conveying control module, a spectral data acquisition and control module and an intelligent recognition and output control module; by arranging the rock debris conveying and preprocessing module, continuous receiving and rapid preparation of rock debris are achieved, meanwhile, the near infrared spectrum detection module conducts spectrum scanning on flowing rock debris above a conveying belt and captures molecular fingerprint information, and generated spectrum data are transmitted to the data processing and intelligent recognition module; the recognition result is synchronously fused with parameters such as drilling depth and pumping pressure, a dynamic slurry diffusion map is generated, the recognition process and the drilling process are kept synchronous, the distribution condition of slurry stone bodies at different depths can be visually seen, and therefore the real-time performance of slurry stone body recognition is effectively improved.

The invention belongs to the technical field of bioinformatics and deep learning, and particularly relates to a DNA methylation site prediction method, system and device and an electronic medium, and the method comprises the following steps: respectively coding single bases in a DNA sequence into four molecular fingerprints, and constructing each base into a molecular map structure; performing high-order feature extraction on the four molecular fingerprints through a multi-layer perceptron to generate molecular fingerprint features; modeling the molecular graph structure by adopting a multi-head graph attention network, calculating and normalizing attention scores among nodes, carrying out weighted summation on neighbor node features, and carrying out multi-head attention fusion to generate molecular graph features; splicing the molecular fingerprint features and molecular map features, and inputting the spliced molecular fingerprint features and molecular map features into a deep convolution gated channel attention module to generate fused features; and obtaining a prediction probability of the fused features, and judging whether the features are methylation sites or not according to the prediction probability. According to the invention, the identification capability of methylation sites can be improved.

The invention relates to the technical field of data processing, in particular to an active cliff data set construction method and device, electronic equipment and a storage medium, and aims to improve the accuracy and consistency of active cliff recognition. The method comprises the following steps: for each molecule which is in a pre-constructed molecular data set, can be combined with the same biological target spot and is subjected to biological activity measurement under the same experimental condition, obtaining the similarity between molecular fingerprints of every two molecules in each molecule, and analyzing the similarity of every two molecules on substructures to obtain the bioactivity of the molecules. Selecting a matching molecule pair meeting a preset similarity evaluation condition; for each biological target spot, selecting a similarity as a molecular activity cliff threshold value corresponding to the biological target spot according to distribution conditions of the corresponding matched molecule pairs under different similarities, and selecting a target molecule pair from the molecules according to the molecular activity cliff threshold value; and constructing an activity cliff data set based on the biological activity parameter difference corresponding to each selected target molecule pair.

This invention provides a system and program that facilitates the process of generating and / or visualizing molecular networks for library molecular structure content. [Solution] In system 500, the molecular network generation system 502 for library molecular structure content includes an evaluation component 512 that performs a comparison between a first molecular fingerprint including first molecular structure data of a first molecular structure and a second molecular fingerprint including second molecular structure data of a second molecular structure, and a visualization component 516 that generates display data that visualizes a representation of the structural similarity score obtained from the first molecular structure, the second molecular structure and the comparison. The representation by the visualization component includes edges corresponding to the structural similarity score that extend between pairs of nodes corresponding to the first molecular structure and the second molecular structure.

The invention provides an ISSR-PCR reaction system for identifying kiwi fruit varieties with different colors and a method and application of the ISSR-PCR reaction system, and belongs to the technical field of molecular markers. The ISSR-PCR reaction system for identifying the kiwi fruit varieties with different colors is as follows: the DNA template is 70ng / mu L, the primer is 1.0 mu mol / L, the Mg < 2 + > is 2mM, the mix is 10mu L, and the remaining ddH2O is prepared into 20 mu L of the system. The amplification procedure is as follows: the pre-denaturation temperature is 94 DEG C, and the time is 5 minutes; the denaturation temperature is 94 DEG C for 30 seconds; the annealing temperature is set according to different primer conditions, and the time is 50s; the extension temperature is 72 DEG C, and the extension time is 1 min; circulation is performed for 35 times; and then extending at 72 DEG C for 7 minutes, and storing at 10 DEG C. The system is used for amplifying 42 kiwi fruit germplasm resources with different colors, the amplification result shows that the established ISSR-PCR reaction system is good in repeatability and high in stability, and the established ISSR molecular fingerprint spectrum can be used for variety identification.

The compound molecular fingerprint prediction belongs to the technical field of mass spectrum data analysis, and from the perspective of fully mining the implicit information of mass spectrum data, neutral loss information is added behind corresponding ion peaks when a vector representation of mass spectrum data is constructed, natural language processing technology is adopted to learn the relationship between peaks and between peaks and neutral losses in the mass spectrum, a multi-dimensional spectrum vector is constructed, and molecular fingerprints are predicted. Due to the addition of neutral loss of corresponding ion peaks in the spectrum vector, more abundant structural information is contained, and the accuracy of molecular fingerprint prediction can be effectively improved.

The invention discloses a terahertz molecular fingerprint sensor based on an over-coupling metasurface and electronic equipment, and relates to the field of terahertz detection. Comprising fingerprint detection unit arrays which are periodically arranged and have the same structure; each fingerprint detection unit comprises a substrate layer, a first metal layer and a protruding structure. The first metal layer is arranged above the substrate layer and has the same cross sectional area with the substrate layer; the protruding structure is arranged above the first metal layer, and the cross section area of the protruding structure is smaller than that of the first metal layer. When the terahertz molecular fingerprint sensor detects a to-be-detected sample, the target position of the first metal layer and the upper surface of the protruding structure are coated with the to-be-detected sample. Terahertz waves are set to vertically enter the over-coupling metasurface, the coupling state of the terahertz waves and the metasurface is changed by adjusting the size of the protruding structure, and qualitative and quantitative detection is carried out on a to-be-detected sample. According to the invention, only a single device is needed, single detection is carried out through a single incident signal, and broadband fingerprint detection and'one-device double-detection 'of a trace sample to be detected are realized.

The invention relates to the field of optical sensing, in particular to a composite structure SERS detection substrate and a preparation method thereof.The composite structure SERS detection substrate comprises a GaN distributed Bragg reflector structure, a first metal layer and a second metal layer, the first metal layer and the second metal layer are deposited on the GaN distributed Bragg reflector structure, and monodisperse metal particles of the first metal layer are selected from Au, Ag or Cu; the monodisperse metal particles of the second metal layer are selected from Pt or Pb. Based on the plasmon resonance characteristic of a small-size multi-metal structure and the high reflection characteristic of a distributed Bragg reflector, the high-sensitivity optical detection of organic molecules is realized by utilizing the organic molecule fingerprint characteristic of a Raman spectrum, and meanwhile, the plasmon catalytic effect of the multi-metal composite structure is modulated by utilizing the multi-metal composite structure, so that the high-sensitivity optical detection of the organic molecules is realized. The optical sensing device is a novel optical sensing device with a wide application prospect.

The invention relates to the technical field of carbohydrate informatics, in particular to a carbohydrate biomacromolecule property prediction method and system based on a graph neural network and chemical fingerprints. Comprising the following steps: modeling a carbohydrate chain sequence into an undirected graph, regarding monosaccharide and glucosidic bonds as nodes in the graph, introducing a virtual node for storing fingerprint features, extracting molecular fingerprint information from the sequence, constructing two different adjacent matrixes Afull and Aori with virtual node connection and without virtual node connection, and constructing two adjacent matrixes Afree and Aori; the dimension of the fingerprint features is reduced to be consistent with the node features, the features are replaced with molecular fingerprint features by positioning the positions of virtual nodes in the graph, the first three layers transmit Aori to enable a model to learn topological structure information of the graph, and the last layer transmits Afull to achieve fusion of the graph structure and the fingerprints. And pooling and splicing node features obtained after convolution of each layer to obtain final sugar chain representation for predicting different properties of sugar chains. According to the method, the graph structure and chemical fingerprint information of the sugar chain can be effectively combined to obtain more meaningful sugar chain representation.

The application provides a set of methods for constructing antigen-antibody complex mutual recognition interface descriptors (i.e., protein molecular fingerprints), which maximally present the structure and physicochemical characteristics of specific recognition by describing the binding interface of the three-dimensional structure of the antibody-antigen from both sides. Based on the protein molecular fingerprints generated by the set of methods, based on the specific interaction recognition rules of the antigen-antibody, the epitope prediction algorithm specific to the antibody and the virtual screening model of the antibody are designed, and the existing machine learning or deep learning tools are docked, so that the antigen epitope prediction based on the antibody and the high-throughput virtual screening of the antibody based on the specific epitope can be quickly realized.

The invention discloses a drug target affinity prediction method, electronic equipment and a computer readable storage medium, the method comprises the following steps: feature extraction is carried out on input small molecule SMILES and protein sequences, and the small molecule adopts 10 molecular fingerprints of RDK, Topological, MACCS, AtomPair, ECFP4, FCFP4, FCFP6, Avalon, Layered and Pattern to construct mixed fingerprint features; compressing the high-dimensional molecular fingerprint features to vector dimensions consistent with protein features through linear mapping to realize feature balance, and splicing to form a fusion feature vector; and performing nonlinear interaction and expression enhancement on the fusion features by adopting a multi-expert hybrid module, and finally outputting an affinity prediction value between the small molecule and the protein through a linear regression layer. Through fusion of multi-source chemical fingerprints and deep protein pre-training representation, higher feature expression ability and stronger model generalization are realized; the complexity of the model is reduced through feature mapping and a lightweight multi-expert hybrid model, so that the model has better stability and expandability.

The application provides a cell-targeting drug primary screening model training method, a drug primary screening method and equipment. The method comprises the following steps: obtaining molecular fingerprint feature data corresponding to each compound sample with label information to form a corresponding original data set, performing data screening on the original data set by using a distance measurement method, and training a regression model based on a graph neural network to learn the structure topology of each compound sample. The model is trained as a cell-targeting drug primary screening model for predicting whether a compound has potential cell-targeting drug activity. The application can effectively improve the comprehensiveness of molecular representation during the training process of the cell-targeting drug primary screening model, effectively improve the generalization ability of the cell-targeting drug primary screening model obtained by training, effectively improve the applicability of the cell-targeting drug primary screening model obtained by training, and improve the accuracy and reliability of the prediction result of the potential cell-targeting drug activity predicted by the model.

The present invention proposes a method and device for intelligently generating molecular fingerprints for non-targeted metabolomics, which relates to the field of data processing technology, including: performing weak enhancement processing on labeled data in mass spectrometry sample data, and performing weak enhancement processing and strong enhancement processing on unlabeled data in mass spectrometry sample data to obtain mass spectrometry training data; using a Transformer encoder to extract features from the mass spectrometry training data to obtain a Morgan molecular fingerprint, and using the Morgan molecular fingerprint as a constraint condition, using the forward diffusion process to generate random Gaussian noise as the basis for reverse diffusion to generate a target molecular fingerprint; using the screened target molecular fingerprint to train a preset network model to obtain a target network model to complete the processing of the mass spectrometry data to be processed. The present invention extracts features from mass spectrometry data and uses them as constraints to generate molecular fingerprints of metabolites to be analyzed, thereby realizing the training of the diffusion model and improving the generalization ability of the model.

The application discloses a molecular odor prediction method based on cross-modal attention fusion, relates to the fields of chemical informatics and deep learning, and comprises the following steps: acquiring a molecular dataset, extracting basic molecular property features, molecular fingerprint features and SMILES sequence features for each molecule in parallel, performing molecular odor prediction model training, wherein the molecular odor prediction model comprises a modal encoder, a cross-modal attention fusion module and a classification head connected in sequence, inputting the corresponding basic molecular property features, molecular fingerprint features and SMILES sequence features of a to-be-predicted molecule into the trained molecular odor prediction model to obtain prediction probabilities of corresponding odor categories, and determining an odor category prediction result of the to-be-predicted molecule based on the prediction probabilities. Through multi-modal feature fusion, an interpretable cross-modal attention and a gating mechanism, the method realizes more reasonable representation modeling of molecular odor, and effectively improves the accuracy of odor label prediction.

The invention provides an organ quality evaluation system and method based on visible-near infrared hyperspectral imaging, and the system employs a continuous gradual change optical filter which can achieve the continuous filtering function from visible light to near infrared light from 400 nm to 1000 nm, and a drive module to form a linear filtering control module, and is used for achieving the filtering of reflected light of a target object (organ) in different wave bands. A reflection spectrum of biological tissues such as hemoglobin and fat is obtained, and molecular fingerprint level data is provided for organ quality evaluation; the imaging camera and the upper computer control module form an imaging processing module, image data of isolated organs of different wave bands are collected through the imaging camera, and spectrum correction, image reconstruction, organ quality evaluation and result display are completed through the upper computer control module. According to the invention, rapid and non-destructive evaluation of the transplanted organ can be realized.

This invention discloses a high-throughput experimental sequence adaptive planning method based on Bayesian optimization, relating to the field of chemical experiment automation technology. The method includes: digitally encoding multiple chemical variables involved in the experiment to construct a multi-dimensional chemical space to be searched; constructing a probabilistic prediction surrogate model based on known experimental samples; pre-setting the acquisition cost of each variable and establishing a cost function; combining the expected improvement amount and the cost function, calculating the next batch of candidate experimental points through a target optimization algorithm, prioritizing the experimental points with the largest unit cost gain; converting the candidate coordinates into instructions and sending them to the automated experimental platform, and updating the surrogate model with the returned experimental data in real time, repeating the process until convergence to the global optimum or reaching the cost ceiling. This invention, by introducing a cost-sensitive acquisition function and a heterogeneous molecular fingerprint fusion strategy, achieves synergistic optimization of experimental results and experimental costs, significantly improving the efficiency and resource utilization of high-throughput experiments.

The invention provides a terahertz metamaterial molecular fingerprint sensing method and system based on multi-task learning, and belongs to the field of terahertz biosensing application. The method comprises the following steps: acquiring a terahertz reflection spectrum of a to-be-detected sample, and extracting physical characteristics of the reflection spectrum; inputting the terahertz reflection spectrum, the corresponding incident angle and the physical characteristics into a trained multi-task learning model, encoding the reflection spectrum, the incident angle and the physical characteristics through a lightweight full-connection network, and mapping the encoded reflection spectrum, the encoded incident angle and the encoded physical characteristics to a unified embedding space; performing feature extraction on the encoded feature vector by using a decoder, and fusing and compressing the feature vector into a compact potential representation through a sharing layer; and inputting the potential representation into a classification head and a regression head, and outputting a molecular category identification result and a concentration prediction result of the to-be-detected sample. According to the method, the performance bottleneck of a traditional single-task model in identification of highly similar molecules is overcome, the problem of low multi-angle scanning efficiency is solved, and high-precision and high-efficiency molecular perception is realized.

The invention provides a Parkinson's disease drug risk prediction method based on knowledge graph embedding and graph neural network, which comprises the following steps: constructing a Parkinson's disease domain knowledge graph containing drugs, targets, metabolic enzymes, adverse reactions and drug interaction relationships, entities and relationships are extracted from authoritative medical guidelines, literatures and databases through a natural language processing technology; based on the knowledge graph, using a knowledge graph embedding model to carry out vectorization expression on drugs and related entities thereof, and combining molecular fingerprint information and protein sequence features of the drugs to construct node features; inputting the node features into a graph neural network model, analyzing a potential action path between drugs through a multi-hop relationship aggregation mechanism, and predicting the occurrence probability of a drug combination and adverse reactions or drug-drug interaction; and outputting the prediction probability, and displaying a knowledge graph path having key influence on a prediction result in a visual mode to assist a doctor in identifying potential drug risks which are not widely reported.

The invention provides a protein-ligand binding affinity prediction method based on a graph convolutional network. The method comprises the following steps: firstly, constructing a protein map based on a three-dimensional structure of a protein pocket, and carrying out feature coding on sequence residues by using ESM3; meanwhile, a molecular map containing geometric information is constructed for the ligand, and chemical characteristics are extracted in combination with molecular fingerprints. And then cross-graph interaction of the protein graph and the ligand graph is realized by adopting a multi-layer message passing graph neural network, and an attention mechanism is introduced to capture key interaction sites and spatial dependency relationships. And generating a unified representation vector of a compound level through pooling, and outputting binding affinity through a prediction network. According to the method, the multi-modal association of the structure and the sequence can be automatically learned, and the accuracy and efficiency of virtual screening and lead compound optimization are effectively improved.

The present invention discloses a method and a screening model for umami peptides. The screening method comprises the following steps: S1: organizing existing umami peptide data and establishing a database; S2: constructing molecular fingerprint feature data of umami peptides based on the structural fragments of the existing umami peptides themselves; S3: constructing intermolecular interaction residue feature data based on the interaction mode between the umami peptides and the umami receptors T1R1 / T1R3 analyzed by molecular docking technology; S4: obtaining molecular descriptor feature data of the physicochemical properties of the umami peptides based on molecular descriptors; S5: using a machine learning algorithm to establish umami peptide screening sub-models for the data obtained in steps S2-S4; S6: integrating the three umami peptide screening sub-models using a support vector machine algorithm to establish an umami peptide screening model; and S7: screening umami peptides using the umami peptide screening model established in step S6. The screening method of the present invention can quickly and accurately screen umami peptides, and the screening method is reusable.

The invention discloses a drug IC50 prediction method and system based on molecular structure and gene expression, and the method comprises the steps: carrying out the comprehensive characterization of the molecular structure of a drug, extracting the chemical structure and characteristic information of the drug through the modes of molecular fingerprints, molecular maps and the like, carrying out the fusion with a gene expression matrix of cells, building a unified characteristic expression system, and carrying out the prediction of the drug IC50. And inferring a gene regulatory network reflecting a potential regulatory relationship between genes based on a variational auto-encoder (VAE). On the basis, a multi-layer feature extraction mechanism combining global and local information is constructed, the global information learns an overall regulation structure among genes in the whole regulation network through a graph neural network, and the local information captures a local action relationship between drugs and key regulation factors by dividing and analyzing sub-graphs. According to the method, the influence mechanism of the drug on the cell system can be described more accurately, the prediction precision of the IC50 value and the interpretability of the model are remarkably improved, and the method has good adaptability and wide application prospects.

The present application relates to a kind of dual-band sensor for mixed substance terahertz molecular fingerprint detection, which is composed of a plurality of molecular fingerprint sensing pixels with different structural parameters, each molecular fingerprint sensing pixel is composed of a plurality of dual-cross aperture unit arrays, the size of dual-cross aperture unit in different molecular fingerprint sensing pixels is different, and each dual-cross aperture unit is composed of two cross aperture unit structures with different arm lengths.The dual-band sensor covers a wider detection range and reduces the number of sensing pixels by using the method of dual-frequency simultaneous scanning, and at the same time, the detection ability of the sensor to trace analysis is enhanced by using surface plasmon polariton, which ensures the significant enhancement of broadband signal of the detected substance.

The invention relates to a regional allergy risk real-time early warning method and device based on an environment micromolecule fingerprint spectrum, and the method comprises the steps: carrying out the real-time sampling of a microvolatile organic compound related to a predefined allergen in the air of a target region, and obtaining an original multi-dimensional characteristic spectrum; extracting a chemical fingerprint spectrum of each known allergen in the original multi-dimensional characteristic spectrum through a VOCs fingerprint spectrum model; dynamically matching the chemical fingerprint spectrum of each known allergen with a preset allergen chemical fingerprint spectrum database by using a Siamese network model, and calculating a comprehensive regional allergy risk factor; and according to the regional allergy risk factor and a preset multi-level allergy risk threshold, determining an allergy risk level of the current target region. According to the method, the purification signal characteristics corresponding to each known allergen are extracted through the VOCs fingerprint spectrum model, and the similarity, relevance and chemical affinity among different allergens are obtained through the Siamese network model, so that the method is closer to the real situation.

The invention belongs to the technical field of crop molecular breeding biology, and particularly provides a soybean whole genome 50K liquid phase chip and application thereof. The soybean whole genome 50K liquid phase chip comprises a soybean 50K site probe mixed solution and a liquid phase hybridization capture reagent, and the soybean 50K site probe mixed solution is composed of 51718 SNP site probes covered by a whole genome; the genome version of the soybean reference genome is Glycine max Williams 82 v 4.0, and the genome version of the soybean reference genome is Glycine max Williams 82 v 4.0; the liquid chip can be used for soybean variety molecular fingerprint construction and rapid authenticity identification, QTL positioning of related traits such as soybean plant type and yield, soybean germplasm resource genetic structure analysis, germplasm resource screening and the like.

The application discloses a kind of based on multimodal data fusion evaluation compound carcinogenic risk deep learning method, device and storage medium, belong to chemical health risk assessment technical field, comprising: (1) establish carcinogenicity prediction dataset;(2) construct cross-modal knowledge graph, adopt convolutional neural network to learn cross-modal knowledge graph and obtain cross-modal feature representation;(3) extract molecular fingerprint feature representation and molecular graph feature representation, obtain fusion feature representation by fusing molecular fingerprint feature representation with cross-modal feature representation;(4) construct fusion model, first deep neural network, second deep neural network in fusion model are used to process different feature representations, and classifier is used for prediction;(5) the parameters of fusion model are optimized by supervision training;(6) the carcinogenic risk of to-be-tested compound is evaluated using the optimized fusion model.The method can comprehensively and accurately evaluate the carcinogenic risk of compound and explain the carcinogenic mechanism.