CN111199779A - Virtual drug screening method and device based on molecular docking - Google Patents
Virtual drug screening method and device based on molecular docking Download PDFInfo
- Publication number
- CN111199779A CN111199779A CN201911363012.XA CN201911363012A CN111199779A CN 111199779 A CN111199779 A CN 111199779A CN 201911363012 A CN201911363012 A CN 201911363012A CN 111199779 A CN111199779 A CN 111199779A
- Authority
- CN
- China
- Prior art keywords
- drug
- information
- network
- atom
- molecular docking
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/50—Molecular design, e.g. of drugs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
- G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
- G16C20/70—Machine learning, data mining or chemometrics
Abstract
The invention provides a virtual drug screening method and a device based on molecular docking, wherein the method comprises the following steps: inputting drug molecule information; converting the drug molecule information into n-dimensional floating point data information; outputting information of each atom of the drug molecules according to n-dimensional floating point type data information by adopting an LSTM network, and summarizing the information of each atom to obtain a vector representing the characteristics of the drug molecules; giving weight to the vector representing the characteristics of the drug molecules through an attention network to obtain a vector model of the drug molecules; the activity of the drug molecules was screened according to the vector model. The scheme of the invention can realize that the AUC value is improved by 2 percent and has stronger robustness.
Description
Technical Field
The invention relates to a virtual drug screening method and a virtual drug screening device based on molecular docking.
Background
The development of new drugs is a very time-consuming, labor-intensive and costly project. In recent years, computer-aided virtual drug screening techniques have been widely developed, and one of the techniques that has been widely used is virtual drug screening based on the principle of molecular docking. The basic principle of the molecular docking technology is to select potential drug molecules to simulate and combine with pathogenic target proteins, but each drug molecule combination mode can be thousands of, and mathematical functions capable of accurately evaluating different combination modes are more difficult to obtain. This is a core problem of current molecular docking techniques.
With the great success of solving the problem in recent years by using an algorithm based on machine learning, such as a support vector machine, a random forest and the like, the performance of a molecular docking model can be improved to a certain extent by using the traditional machine learning technology, but the method obviously limits the wide application of the method by artificially extracting molecular features. In addition, the characteristics acquired by the method have no interpretability on some practical complex problems, and the generalization of the model is poor.
On the other hand, another branch of machine learning, deep learning, is increasingly being applied to this field. The method can automatically and directly extract features from the molecular expression, and the model built by the method is easier to migrate. In 2014, a technology emerged that predicted biochemical activity by establishing a multitask neural network and then according to molecular structure. Immediately after 2015, techniques for predicting pharmaceutical activity based on molecular fingerprinting have emerged.
At present, the deep learning algorithm is less applied to virtual screening based on molecular docking, more applications are applied to optimization of a traditional machine model, and domain experts are required to participate in feature engineering, so that the quality of the model is determined by the quality of the feature engineering. This limits the wide application of this kind of algorithm, and in addition this kind of algorithm is very unstable to the output simulation effect of different molecule butt joint quality.
Disclosure of Invention
Aiming at the problems in the related art, the invention aims to provide a virtual drug screening method and device based on molecular docking. Compared with the result of the latest algorithm on the same training set, the AUC value of the algorithm is improved by 2%, and the algorithm has stronger robustness on the training sets of 40 different pathogenic proteins.
According to an embodiment of the invention, a virtual drug screening method based on molecular docking comprises the following steps: inputting drug molecule information; converting the drug molecule information into n-dimensional floating point data information; outputting information of each atom of the drug molecules according to n-dimensional floating point type data information by adopting an LSTM network, and summarizing the information of each atom to obtain a vector representing the characteristics of the drug molecules; giving weight to the vector representing the characteristics of the drug molecules through an attention network to obtain a vector model of the drug molecules; the activity of the drug molecules was screened according to the vector model.
According to the embodiment of the invention, the virtual drug screening method based on molecular docking comprises the following steps: atom type, chemical bond energy size, atomic spacing, and type of bound amino acid.
According to the embodiment of the invention, the virtual drug screening method based on molecular docking comprises the following steps: and filling the data sequence of the drug molecule information before inputting into the LSTM network so as to ensure that the length of the batch data sequence is consistent.
According to the embodiment of the invention, the virtual drug screening method based on molecular docking comprises the following steps: for any drug molecule x, consisting of m atoms, the matrix representation of the input LSTM is { z1, z2, · · · ·, zm }, where zi is the information vector of the ith atom, and then the output of each atom through the network is:
ui＝f(Wizi+bi)，
f () is the output function of the LSTM network.
According to the embodiment of the invention, the virtual drug screening method based on molecular docking, wherein the step of summarizing and obtaining the vector representing the molecular characteristics of the drug from the information of each atom comprises the following steps: and (3) generating final characteristics by adopting a maximum pooling method, wherein the specific formula is as follows:
according to the embodiment of the invention, the virtual drug screening method based on molecular docking comprises the following steps: the weight calculation is obtained by 2-layer nonlinear transformation, and is used as each vector r representing the molecular characteristics of the drugiA weight w of 0 to 1iCorrected feature vector
The calculation formula of (2) is as follows:
according to the embodiment of the invention, the virtual drug screening method based on molecular docking further comprises the step that the attention network comprises a first attention network and a second attention network, the first attention network adopts a relu activation function to screen effective vectors, and the second attention network adopts a sigmoid activation function to enable the weight to be within the range of 0-1.
According to the embodiment of the invention, the virtual drug screening method based on molecular docking comprises the following steps: weight wiThe calculation formula is as follows:
wherein W1、b1Is the weight and bias of the first attention network, W2、b2As weights and biases for the second attention network, first notesThe number of layers of the attention network is j/r, r is a set hyper-parameter of compressed data, and the number of layers of the second attention network is j.
According to an embodiment of the present invention, a virtual drug screening device based on molecular docking includes: the input module is used for inputting drug molecule information; the data conversion module is used for converting the drug molecule information into n-dimensional floating point data information; the LSTM network module is used for outputting information of each atom of the medicine molecules according to n-dimensional floating point type data information by adopting an LSTM network, and summarizing the information of each atom to obtain a vector representing the characteristics of the medicine molecules; the attention network module is used for giving weight to the vector representing the characteristics of the drug molecules through an attention network to obtain a vector model of the drug molecules; and the screening module is used for screening the activity of the drug molecules according to the vector model.
According to an embodiment of the present invention, a virtual drug screening device based on molecular docking includes: the drug molecule information includes: atom type, chemical bond energy size, atomic spacing, and type of bound amino acid.
The invention has the beneficial technical effects that: the scheme of the invention omits the feature engineering construction of human participation, expands the structural information into n-dimensional vectors by using an Embedding method, and automatically extracts features through an LSTM network and an attention mechanism. Compared with the result of the latest algorithm on the same training set, the AUC value of the algorithm is improved by 2%, and the algorithm has stronger robustness on the training sets of 40 different pathogenic proteins.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a flow chart of a method for virtual drug screening based on molecular docking, according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of the general architecture of a virtual drug screening method based on molecular docking according to an embodiment of the present invention;
FIG. 3 is a schematic representation of the conversion of drug molecule and pathogenic protein complex information to n-scale structural information according to one embodiment of the present invention;
FIG. 4 is a schematic diagram of an LSTM layer network structure according to one embodiment of the invention;
fig. 5 is a schematic diagram of an attention layer network according to one embodiment of the invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments that can be derived by one of ordinary skill in the art from the embodiments given herein are intended to be within the scope of the present invention.
As shown in fig. 1, the present invention provides a virtual drug screening method based on molecular docking, comprising:
and S11, inputting the information of the drug molecules.
And S12, converting the drug molecule information into floating point type data information with n dimensions (n is a positive integer).
And S13, outputting information of each atom of the drug molecules according to n-dimensional floating point type data information by adopting an LSTM (Long Short-Term Memory) network, and summarizing the information of each atom to obtain a vector representing the characteristics of the drug molecules.
And S14, weighting the vectors representing the characteristics of the drug molecules through the attention network to obtain a vector model of the drug molecules.
And S15, screening the activity of the drug molecules according to the vector model.
According to the technical scheme, the structural information is expanded into n-dimensional vectors, and features are automatically extracted by using an LSTM network and an attention mechanism, so that the screening of the drug molecule activity is realized. The technical scheme of the invention has the advantages that: 1. the model is highly automatic, and the characteristic engineering construction without human participation is carried out in the whole process. 2. For the molecular docking output with different masses, compared with the prior art, the algorithm model has strong robustness. 3. Compared with the prior art, the AUC value overall mean value of the algorithm is improved by 2% under the same experimental conditions when 40 different protein data sets are tested.
The general architecture of the algorithmic model of the present invention is shown in FIG. 2. In order to perform virtual drug screening on molecular docking data with different qualities, the scheme of the invention adopts a SMILES mode as an input text for a molecular structure, performs data coding through an Embedding layer, acquires dependence information inside the molecular structure in an LSTM network, acquires structural characteristics through piece pooling, and finally outputs a prediction result through a softmax layer.
1. SMILES mode representation
In the scheme of the invention, the molecular structure input comprises 4 parts of information: atom type, chemical bond energy size, atom spacing, type of bound amino acid. Since the invention predicts the activity of the drug molecule against the pathogenic protein, the inputted molecular structure not only has the information of the internal structure of the molecule, but also contains the type of amino acid bound by the receptor molecule. The input is converted into digital structural information after data preprocessing and is input into a network.
2. Embedding layer
The Embedding layer is used for converting structural information represented by long integer numbers into floating point data information with n dimensions. The dimension n is used as a network hyper-parameter and can be obtained by parameter adjustment. As shown in FIG. 3, the drug molecule and pathogenic protein complex is expressed by SMILES and then converted into n-scale structural information through the Embedding layer.
3. LSTM layer
In order to fully extract the structural characteristics of molecules and obtain effective associated information of the internal atomic structures of the molecules, the invention adopts an LSTM network. The LSTM layer network structure is shown in figure 4. Because the lengths of the drug molecules are different, the batch sequences are filled before network input so as to ensure that the lengths of the batch sequences are consistent. The network can be expressed by the following mathematical formula, and for any drug molecule x, the matrix of the input LSTM can be expressed as { z1, z2 }, zm }. Where zi is the information vector of the ith atom. The output of each atom through the network is then:
ui＝f(Wizi+bi)，
where f () is the output function of the LSTM network.
Next, the second step is a pooling layer, aggregating features from different atomic information. The final features are generated by a method of maximum pooling. The specific formula is as follows:
in this way, the newly formed vector r represents all structural information of the drug molecule-pathogenic protein complex, and then through continuous learning and training, the network can finally obtain the key structural information for distinguishing whether the drug molecule is effective or not.
4. Attention layer
The attention layer network is shown in fig. 5. After the vectors r pass through the attention layer, each vector riA weight w of 0 to 1iThe weight calculation is obtained by a 2-layer nonlinear transformation. Modified feature vector
Calculating the formula:
weight wiCalculating the formula:
Wi＝Sigmoid(W2(Relu(W2r+b1))+b2)
wherein W1，b1，W2，b2The weights and biases for attention network 1 and network 2, respectively. The network 1 adopts a relu activation function to screen effective vectors, and the network 2 adopts a sigmoid activation function to enable the weight to be within a range of 0-1. The network is as shown in the figureThe number of layers of the network 1 is shown as the number j/r of input r, r is a hyper-parameter which needs to be set, and the purpose is to compress data and obtain effective characteristics. The number of layers of the network 2 is equal to j, which ensures consistency with the input.
5. Scoring system
Corrected vector
After 2 layers of conventional hidden layer processing, the last layer is calculated by a softmax function to obtain 2 classified result scores: 0 represents no activity of the drug molecule and 1 represents active. Is formulated as follows, with the final output s for any compound x(x)Comprises the following steps:
s(x)＝W3(W2r+b2)+b3
s(x)is a vector matrix of 2 dimensions and is,
and
class 0 and 1 possibilities, respectively. The probability is used for calculation through a Softmax function, and the specific formula is as follows:
in the network training phase, the algorithm employs Adam gradient descent to minimize the loss function loss. loss adopts an L2 loss function, and the calculation formula is as follows:
where θ is the vector matrix of all parameters W1,.. The parameters x and y respectively correspond to the structureThe input information and the corresponding genuine label classification 0 or 1. The first half of the formula is used to calculate the error from the true value, and the second half prevents the model from being over-fitted.
During the experiment, a batch training method was used, with a batch size of 20. The verification method is 1-fold cross verification.
The invention also provides a virtual drug screening device based on molecular docking, which comprises: the input module is used for inputting drug molecule information; the data conversion module is used for converting the drug molecule information into n-dimensional floating point data information; the LSTM network module is used for outputting information of each atom of the medicine molecules according to n-dimensional floating point type data information by adopting an LSTM network, and summarizing the information of each atom to obtain a vector representing the characteristics of the medicine molecules; the attention network module is used for giving weight to the vector representing the characteristics of the drug molecules through an attention network to obtain a vector model of the drug molecules; and the screening module is used for screening the activity of the drug molecules according to the vector model.
In one embodiment, the above virtual drug screening apparatus based on molecular docking includes: atom type, chemical bond energy size, atomic spacing, and type of bound amino acid.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes will occur to those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A virtual drug screening method based on molecular docking is characterized by comprising the following steps:
inputting drug molecule information;
converting the drug molecule information into n-dimensional floating point data information;
outputting information of each atom of the medicine molecules according to the n-dimensional floating point type data information by adopting a long-short term memory (LSTM) network, and summarizing the information of each atom to obtain a vector representing the characteristics of the medicine molecules;
giving weight to the vector representing the characteristics of the drug molecules through an attention network to obtain a vector model of the drug molecules;
and screening the activity of the drug molecules according to the vector model.
2. The virtual drug screening method based on molecular docking of claim 1, wherein the drug molecular information comprises: atom type, chemical bond energy size, atomic spacing, and type of bound amino acid.
3. The virtual drug screening method based on molecular docking as claimed in claim 1, comprising: and filling the data sequence of the drug molecule information before inputting into the LSTM network so as to ensure that the length of the batch data sequence is consistent.
4. The virtual drug screening method based on molecular docking as claimed in claim 3, comprising: for any drug molecule x, consisting of m atoms, the matrix representation of the input LSTM is { z1, z2,. cndot., zm }, where zi is the information vector of the ith atom, Wi、biFor the weight and bias of the LSTM network, the output of each atom through the network is:
ui＝f(Wizi+bi)，
where f () is the output function of the LSTM network.
5. The virtual drug screening method based on molecular docking of claim 4, wherein the aggregating the vector representing the molecular characteristics of the drug from the information of each atom comprises: and (3) generating final characteristics by adopting a maximum pooling method, wherein the specific formula is as follows:
6. the virtual drug screening method based on molecular docking as claimed in claim 1, comprising: the weight calculation is obtained by 2-layer nonlinear transformation and is used for each vector r representing the molecular characteristics of the medicineiA weight w of 0 to 1iCorrected feature vector
The calculation formula of (2) is as follows:
7. the virtual drug screening method based on molecular docking of claim 6, further comprising the attention network comprises a first attention network and a second attention network, the first attention network adopts a relu activation function to screen the valid vectors, and the second attention network adopts a sigmoid activation function so that the weight is within an interval of 0-1.
8. The virtual drug screening method based on molecular docking as claimed in claim 7, comprising: the weight wiThe calculation formula is as follows:
Wi＝Sigmoid(W2(Relu(W1r+b1))+b2)
wherein W1、b1Is the weight and bias of the first attention network, W2、b2The number of layers of the first attention network is j/r, r is a set hyper-parameter of compressed data, and the number of layers of the second attention network is j.
9. A virtual drug screening device based on molecular docking, comprising:
the input module is used for inputting drug molecule information;
the data conversion module is used for converting the drug molecule information into n-dimensional floating point data information;
the LSTM network module is used for outputting information of each atom of the medicine molecules according to the n-dimensional floating point type data information by adopting an LSTM network, and summarizing the information of each atom to obtain a vector representing the characteristics of the medicine molecules;
the attention network module is used for giving weight to the vector representing the drug molecule characteristics through an attention network to obtain a vector model of the drug molecules;
and the screening module is used for screening the activity of the drug molecules according to the vector model.
10. The virtual drug screening device based on molecular docking as claimed in claim 9, comprising: the drug molecule information includes: atom type, chemical bond energy size, atomic spacing, and type of bound amino acid.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911363012.XA CN111199779A (en) | 2019-12-26 | 2019-12-26 | Virtual drug screening method and device based on molecular docking |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911363012.XA CN111199779A (en) | 2019-12-26 | 2019-12-26 | Virtual drug screening method and device based on molecular docking |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111199779A true CN111199779A (en) | 2020-05-26 |
Family
ID=70744414
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911363012.XA Pending CN111199779A (en) | 2019-12-26 | 2019-12-26 | Virtual drug screening method and device based on molecular docking |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111199779A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113436686A (en) * | 2021-06-23 | 2021-09-24 | 腾讯科技（深圳）有限公司 | Artificial intelligence-based compound library construction method, device, equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107862173A (en) * | 2017-11-15 | 2018-03-30 | 南京邮电大学 | A kind of lead compound virtual screening method and device |
| CN109033738A (en) * | 2018-07-09 | 2018-12-18 | 湖南大学 | A kind of pharmaceutical activity prediction technique based on deep learning |
| CN110459274A (en) * | 2019-08-01 | 2019-11-15 | 南京邮电大学 | A kind of small-molecule drug virtual screening method and its application based on depth migration study |
| CN110534164A (en) * | 2019-09-26 | 2019-12-03 | 广州费米子科技有限责任公司 | Drug molecule generation method based on deep learning |
-
2019
- 2019-12-26 CN CN201911363012.XA patent/CN111199779A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107862173A (en) * | 2017-11-15 | 2018-03-30 | 南京邮电大学 | A kind of lead compound virtual screening method and device |
| CN109033738A (en) * | 2018-07-09 | 2018-12-18 | 湖南大学 | A kind of pharmaceutical activity prediction technique based on deep learning |
| CN110459274A (en) * | 2019-08-01 | 2019-11-15 | 南京邮电大学 | A kind of small-molecule drug virtual screening method and its application based on depth migration study |
| CN110534164A (en) * | 2019-09-26 | 2019-12-03 | 广州费米子科技有限责任公司 | Drug molecule generation method based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| SHUANGJIA ZHENG 等: ""Identifying Structure−Property Relationships through SMILES Syntax Analysis with Self-Attention Mechanism"" * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113436686A (en) * | 2021-06-23 | 2021-09-24 | 腾讯科技（深圳）有限公司 | Artificial intelligence-based compound library construction method, device, equipment and storage medium |
| WO2022267752A1 (en) * | 2021-06-23 | 2022-12-29 | 腾讯科技（深圳）有限公司 | Compound processing method and apparatus based on artificial intelligence, and device, storage medium, and computer program product |
| CN113436686B (en) * | 2021-06-23 | 2024-02-27 | 腾讯科技（深圳）有限公司 | Method, device, equipment and storage medium for constructing compound library based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7312906B2 (en) | Resource Constrained Neural Network Architecture Search | |
| US11100403B2 (en) | Cooperative execution of a genetic algorithm with an efficient training algorithm for data-driven model creation | |
| KR102302609B1 (en) | Neural Network Architecture Optimization | |
| US20190130277A1 (en) | Ensembling of neural network models | |
| CN112287982A (en) | Data prediction method and device and terminal equipment | |
| CN111602148A (en) | Regularized neural network architecture search | |
| CN109376857A (en) | A kind of multi-modal depth internet startup disk method of fusion structure and attribute information | |
| CN113140254B (en) | Meta-learning drug-target interaction prediction system and prediction method | |
| CN111429977B (en) | Novel molecular similarity search algorithm based on attention of graph structure | |
| Chen et al. | Improving the training and evaluation efficiency of recurrent neural network language models | |
| CN111723914A (en) | Neural network architecture searching method based on convolution kernel prediction | |
| CN110349597A (en) | A kind of speech detection method and device | |
| CN114491039B (en) | Primitive learning few-sample text classification method based on gradient improvement | |
| CN112017255A (en) | Method for generating food image according to recipe | |
| Ding et al. | Conditional generative adversarial network model for simulating intensity measures of aftershocks | |
| US20220383119A1 (en) | Granular neural network architecture search over low-level primitives | |
| Rajapakshe et al. | A novel policy for pre-trained deep reinforcement learning for speech emotion recognition | |
| Huai et al. | Latency-constrained DNN architecture learning for edge systems using zerorized batch normalization | |
| CN111199779A (en) | Virtual drug screening method and device based on molecular docking | |
| CN113611354A (en) | Protein torsion angle prediction method based on lightweight deep convolutional network | |
| KR20220066554A (en) | Method, apparatus and computer program for buildding knowledge graph using qa model | |
| CN116825236A (en) | Method, device, equipment and medium for generating drug molecules of protein targets | |
| KR20190082453A (en) | Method, apparatus and computer program for analyzing new learning contents for machine learning modeling | |
| CN114742292A (en) | Knowledge tracking process-oriented two-state co-evolution method for predicting future performance of students | |
| CN113112969A (en) | Buddhism music score recording method, device, equipment and medium based on neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |