Method, device and equipment for generating solvate prediction model and application thereof
By training a neural network model to generate a solvate prediction model, the problem of time-consuming and labor-intensive solvate screening in drug development has been solved, enabling rapid and low-cost solvate recommendation and improving screening efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN JINGTAI TECH CO LTD
- Filing Date
- 2022-12-30
- Publication Date
- 2026-05-01
AI Technical Summary
Existing technologies for screening solvates in drug development rely on experiments that consume a lot of materials, manpower, and time, or depend too much on expert experience, making it impossible to design solutions with high throughput.
By acquiring training data, an initial neural network model is trained to generate a solvate prediction model. This model is then used to quickly and cost-effectively recommend solvent molecules that can generate solvates, including feature extraction of small molecule compounds and solvent molecules represented using SMILES. Feature extraction and prediction are performed by combining graph neural networks and fully connected neural networks.
This technology enables efficient, rapid, and low-cost recommendation of solvent molecules that can form solvates for small molecule compounds, reducing experimental costs and time and improving the efficiency of solvate screening.
Smart Images
Figure CN116052800B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the pharmaceutical field, and more particularly to a method, apparatus, and device for generating a solvate prediction model, and a method, apparatus, device, and computer-readable storage medium for predicting solvates based on the solvate prediction model. Background Technology
[0002] Currently, most drugs on the market exist in crystalline form. Different solid forms can be obtained during drug crystallization, such as polymorphs, hydrates, solvates, salts, and eutectics. In new drug development and manufacturing, methods such as precipitation, crystallization, and recrystallization are commonly used. The key to these methods is selecting a suitable solvent for experimentation. During the contact between drug molecules and solvent molecules, crystals composed of both drug and solvent molecules may form, known as solvates.
[0003] Solvates were initially obtained accidentally during the crystallization process of separation and purification. With increasing attention paid to the polymorphism of drugs, drug developers have actively sought more polymorphic substances and solvates through extensive experiments using various solvents and experimental methods. A wide variety of solvents are available. Current solvent screening methods primarily involve extensively screening candidate solvent molecules for specific drug molecules using numerous experimental methods to identify the appropriate solvents to form solvates. Alternatively, relying on expert experience, experimental experts, through extensive study and practice, possess a deep understanding of the crystallization characteristics of different types of drug molecules and can recommend experimental solvents for specific drug molecules, thus determining the solvents that can form solvates with a small number of experiments.
[0004] However, methods that screen candidate solvent molecules require too many experiments, resulting in high costs in terms of materials, manpower, and time. On the other hand, methods that rely on expert experience are too dependent on personal experience and cannot design solutions with high throughput. Summary of the Invention
[0005] This application provides a method, apparatus, and device for generating a solvate prediction model, as well as a solvate prediction method, apparatus, and device based on the solvate prediction model, and a computer-readable storage medium, which can efficiently, rapidly, and cost-effectively recommend solvent molecules that can form solvates for a given small molecule compound. The method for generating the solvate prediction model includes:
[0006] Acquire training data, which includes positive samples and negative samples. The positive samples include multiple small molecule compounds and information on solvent molecules that can form solvates with the multiple small molecule compounds. The negative samples include multiple small molecule compounds and information on solvent molecules that cannot form solvates with the multiple small molecule compounds.
[0007] An initial neural network model is trained based on the training data to obtain a solvate prediction model.
[0008] Optionally, the small molecule compounds in the positive and / or negative samples are represented by SMILES; and / or,
[0009] The solvates in the positive and / or negative samples are represented by SMILES.
[0010] Optionally, the initial neural network model includes a first graph neural network, a second graph neural network, and a fully connected neural network; wherein,
[0011] The first graph neural network is used to extract features of small molecule compounds in the positive samples and features of small molecule compounds in the negative samples;
[0012] The second graph neural network is used to extract features of solvent molecules in the positive samples and features of solvent molecules in the negative samples;
[0013] The fully connected neural network is used to extract feature vectors that connect the features of the small molecule compound and the features of the solvent molecule, and outputs the probability of a specific small molecule compound forming a solvate with different solvent molecules based on the feature vectors.
[0014] Optionally, the probability that the specific small molecule compound will form a solvate with different solvent molecules is output in the form of a score.
[0015] Optionally, acquiring training data includes:
[0016] Training data can be collected from at least one of the following sources: papers, databases, and experimental data on solvates.
[0017] This method for predicting solubilities based on a solubility prediction model includes:
[0018] Obtain a predictive model for solvates;
[0019] The solvate prediction model takes a small molecule compound to be predicted as input, and predicts recommended solvent molecules that may form solvates with the small molecule compound.
[0020] The solvate prediction model is input with the small molecule compound to be predicted and the candidate solvent molecule, and the probability of the small molecule compound to be predicted forming a solvate with the candidate solvent molecule is predicted by the solvate prediction model.
[0021] Optionally, the solvate prediction model includes a first graph neural network, a second graph neural network, and a fully connected neural network; wherein,
[0022] When the solvate prediction model is input with a small molecule compound to be predicted and a candidate solvent molecule, the step of predicting the probability of the small molecule compound to be predicted forming a solvate with the candidate solvent molecule through the solvate prediction model includes:
[0023] The features of the small molecule compound to be predicted are extracted using the first graph neural network;
[0024] The features of the candidate solvent molecules are extracted using the second graph neural network;
[0025] The fully connected neural network extracts feature vectors that connect the features of the small molecule compound to be predicted and the features of the candidate solvent molecule, and outputs the probability that the small molecule compound to be predicted will form a solvate with the candidate solvent molecule based on the feature vectors.
[0026] Optionally, the solvate prediction model includes a first graph neural network, a second graph neural network, and a fully connected neural network; wherein,
[0027] When the solvate prediction model is input with a small molecule compound to be predicted, the step of predicting recommended solvent molecules that may form solvates with the small molecule compound through the solvate prediction model includes:
[0028] The features of the small molecule compound to be predicted are extracted using the first graph neural network;
[0029] The second graph neural network provides features for a variety of candidate solvent molecules;
[0030] The fully connected neural network extracts feature vectors that connect the features of the small molecule compound to be predicted and the features of the multiple candidate solvent molecules, and obtains the probability that the small molecule compound to be predicted will form solvates with the multiple candidate solvent molecules based on the feature vectors.
[0031] Based on the stated probability, recommended solvent molecules are obtained that may form solvates with the predicted small molecule compounds.
[0032] The apparatus for generating the solvate prediction model includes a memory and a processor, wherein the memory stores executable code that, when processed by the processor, can cause the processor to execute any of the methods described above.
[0033] The solvate prediction device based on the solvate prediction model includes a memory and a processor. The memory stores executable code, which, when processed by the processor, can cause the processor to perform any of the methods described above.
[0034] The apparatus for obtaining a solvate prediction model includes:
[0035] An acquisition module is used to acquire training data, which includes positive samples and negative samples. The positive samples include multiple small molecule compounds and information on solvent molecules that can form solvates with the multiple small molecule compounds. The negative samples include multiple small molecule compounds and information on solvent molecules that cannot form solvates with the multiple small molecule compounds.
[0036] The training module is used to train an initial neural network model based on the training data to obtain a solvate prediction model.
[0037] The solvate prediction device based on the solvate prediction model includes:
[0038] The acquisition module is used to acquire the solvate prediction model;
[0039] The solvate prediction module is used to input the small molecule compound to be predicted into the solvate prediction model, and predict recommended solvent molecules that may form solvates with the small molecule compound to be predicted, based on the solvate prediction model; or, it is used to input the small molecule compound to be predicted and candidate solvent molecules into the solvate prediction model, and predict the probability that the small molecule compound to be predicted and the candidate solvent molecules will form solvates based on the solvate prediction model.
[0040] The computer-readable storage medium stores executable code that, when executed by a processor of an electronic device, causes the electronic device to perform any of the methods described above.
[0041] In this embodiment, a solvate prediction model is obtained by training a neural network with collected training data. This solvate prediction model can quickly and effectively recommend solvent molecules that can generate solvates for a given small molecule compound. Attached Figure Description
[0042] Figure 1 This is a schematic diagram of an embodiment of the method for generating a solvate prediction model according to this application;
[0043] Figure 2 This is a schematic diagram of an embodiment of the solvate prediction model in this application;
[0044] Figure 3This is a schematic diagram of an embodiment of the solvate prediction method based on a solvate prediction model in this application;
[0045] Figure 4 This is a schematic diagram of an embodiment of the apparatus for generating the solvate prediction model of this application;
[0046] Figure 5 This is a schematic diagram of an embodiment of the solvate prediction device based on the solvate prediction model in this application;
[0047] Figure 6 This is a schematic diagram of an embodiment of the apparatus for generating a solvate prediction model according to this application;
[0048] Figure 7 This is a schematic diagram of an embodiment of the solvate prediction device based on the solvate prediction model in this application. Detailed Implementation
[0049] Embodiments of this application will now be described in more detail with reference to the accompanying drawings. While embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to make this application more thorough and complete, and to fully convey the scope of this application to those skilled in the art.
[0050] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.
[0051] It should be understood that although the terms "first," "second," "third," etc., may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0052] like Figure 1 The above, Figure 1 This is a schematic diagram of an embodiment of the method for generating a solvate prediction model according to this application. The method for generating the solvate prediction model includes:
[0053] Step S101: Obtain training data.
[0054] In this example, the training data includes positive and negative samples. The positive samples include multiple small molecule compounds and information on solvent molecules that can form solvates with each of the small molecule compounds. The negative samples include multiple small molecule compounds and information on solvent molecules that cannot form solvates with each of the small molecule compounds. Optionally, the small molecule compounds in the positive and / or negative samples are represented using the Simplified Molecular Input Line Entry System (SMILES). Optionally, the solvates in the positive and / or negative samples are represented using SMILES.
[0055] Optionally, training data can be obtained from at least one of the following sources: publicly published papers, databases, and experimental data on solvates. After collecting this data from these sources, SMILES of multiple small molecule compounds and SMILES of solvent molecules that can form solvates with the multiple small molecule compounds are extracted as positive sample inputs, and SMILES of multiple small molecule compounds and SMILES of solvent molecules that cannot form solvates with the multiple small molecule compounds are extracted as negative sample inputs.
[0056] Step S102: Train an initial neural network model based on the training data to obtain a solvate prediction model.
[0057] There are various methods for training the initial neural network model, such as using existing training methods. The initial neural network model can be of various types, such as... Figure 2 As shown, Figure 2 This is a schematic diagram of one embodiment of the initial neural network model in this application. The initial neural network model 20 includes a first graph neural network 21, a second graph neural network 22, and a fully connected neural network 23.
[0058] The first graph neural network 21 is used to extract features of small molecule compounds in the positive samples and features of small molecule compounds in the negative samples. The second graph neural network 22 is used to extract features of solvent molecules in the positive samples and features of solvent molecules in the negative samples. The fully connected neural network 23 is used to extract feature vectors connecting the features of the small molecule compounds and the features of the solvent molecules, and outputs the probability of a specific small molecule compound forming a solvate with different solvent molecules based on the feature vectors.
[0059] Optionally, the probability of a specific small molecule compound forming a solvate with different solvent molecules is output as a score. Optionally, this probability is expressed as a score, indicating whether the input drug molecule and the input solvent molecule are likely to form a solvate. For example, the score ranges from 0 to 1, with a higher value indicating a greater likelihood of the drug molecule and solvent forming a solvate, and a lower value indicating a greater likelihood of the drug molecule and solvent not forming a solvate.
[0060] like Figure 3 The above, Figure 3 This is a schematic diagram of an embodiment of the solvate prediction method based on a solvate prediction model according to this application. The solvate prediction method based on the solvate prediction model includes:
[0061] Step S301: Obtain the solvate prediction model.
[0062] The solvate prediction model can be a solvate prediction model generated according to the above-described method. For example, the solvate prediction model includes a first graph neural network, a second graph neural network, and a fully connected neural network.
[0063] Step S302: Input the small molecule compound to be predicted into the solvate prediction model, and use the solvate prediction model to predict recommended solvent molecules that may form solvates with the small molecule compound to be predicted, or...
[0064] Step S303: Input the small molecule compound to be predicted and the candidate solvent molecule into the solvate prediction model, and predict the probability of the small molecule compound to be predicted forming a solvate with the candidate solvent molecule through the solvate prediction model.
[0065] In the case where the solvate prediction model includes a first graph neural network, a second graph neural network, and a fully connected neural network, optionally, in step S302, after inputting the small molecule compound to be predicted and the candidate solvent molecule into the solvate prediction model, the features of the small molecule compound to be predicted can be extracted by the first graph neural network; the features of the candidate solvent molecule can be extracted by the second graph neural network; the feature vector connecting the features of the small molecule compound to be predicted and the features of the candidate solvent molecule can be extracted by the fully connected neural network; and the probability of the small molecule compound to be predicted forming a solvate with the candidate solvent molecule can be output based on the feature vector.
[0066] Optionally, this probability can be determined by a score indicating whether the predicted small molecule compound and the candidate solvent molecule are likely to form a solvate. For example, the score ranges from 0 to 1, with a higher value indicating a greater likelihood of solvate formation between the drug molecule and the solvent, and a lower value indicating a greater likelihood of non-solvate formation.
[0067] In cases where the solvate prediction model includes a first graph neural network, a second graph neural network, and a fully connected neural network, optionally, in step S303, after inputting the small molecule compound to be predicted into the solvate prediction model, if no candidate solvent molecules are input, common solvent molecules are used as candidate solvent molecules, or all solvent molecules are used as candidate solvent molecules, or solvent molecules in the training data are used as candidate solvent molecules. Features of the small molecule compound to be predicted are extracted using the first graph neural network; features of candidate solvent molecules are provided using the second graph neural network; feature vectors connecting the features of the small molecule compound to be predicted and the features of different candidate solvent molecules are extracted using the fully connected neural network; and the probability of the small molecule compound to form solvates with different candidate solvent molecules is obtained based on the feature vectors; and recommended solvent molecules that may form solvates with the small molecule compound to be predicted are obtained based on the probability.
[0068] Optionally, when obtaining recommended solvent molecules that may form solvates with the predicted small molecule compound based on the probability, the probability of the predicted small molecule compound forming solvates with different candidate solvent molecules can be sorted by magnitude, and candidate solvent molecules with a probability greater than a preset value or falling within a preset range are selected as recommended solvent molecules. Optionally, the output result also includes the probability of the recommended solvent molecule forming a solvate with the predicted small molecule compound.
[0069] In a specific example, using axitinib as the drug molecule, the SMILES of axitinib are input into the prediction model:
[0070] CNC(=O)C1=CC=CC=C1SC2=CC3=C(C=C2)C(=NN3) / C=C / C4=CC=CC=N4. The solvate prediction model is also fed a set of candidate solvent molecules, including methanol, ethanol, acetone, chloroform, and ethyl acetate. The prediction model outputs a score and ranking indicating the probability of each candidate solvent molecule set and axitinib forming a solvate, as shown in the table below:
[0071] solvent Fraction Sort methanol 0.945 1 ethanol 0.922 2 acetone 0.881 3 chloroform 0.876 4 Ethyl acetate 0.891 5
[0072] In this embodiment, a solvate prediction model is obtained by training an initial neural network model with collected training data. This solvate prediction model can be used to quickly and effectively predict the probability of a given small molecule compound and multiple candidate solvent molecules forming a solvate, or to efficiently, quickly, and cost-effectively recommend solvent molecules that can form a solvate for a given small molecule compound.
[0073] like Figure 4 As shown, this application embodiment also provides a solvate prediction model generation device 40, including a first memory 41 and a first processor 42. The first memory 41 stores executable code. When the executable code is processed by the first processor 42, the first processor 42 can execute the above-described solvate prediction model generation method.
[0074] like Figure 5 As shown in the embodiment of this application, a solvate prediction device 50 based on a solvate prediction model is also provided, including a second memory 51 and a second processor 52. The second memory 51 stores executable code. When the executable code is processed by the second processor 52, the second processor 52 can execute the above-described solvate prediction method based on the solvate prediction model.
[0075] like Figure 6 As shown in the embodiments of this application, an apparatus 60 for generating a solvate prediction model is also provided, comprising:
[0076] The acquisition module 61 is used to acquire training data, which includes positive samples and negative samples. The positive samples include multiple small molecule compounds and information on solvent molecules that can form solvates with the multiple small molecule compounds. The negative samples include multiple small molecule compounds and information on solvent molecules that cannot form solvates with the multiple small molecule compounds.
[0077] Training module 62 is used to train an initial neural network model based on the training data to obtain a solvate prediction model.
[0078] Optionally, the small molecule compounds in the positive and / or negative samples are represented by SMILES; and / or,
[0079] The solvates in the positive and / or negative samples are represented by SMILES.
[0080] Optionally, the initial neural network model includes a first graph neural network, a second graph neural network, and a fully connected neural network; wherein,
[0081] The first graph neural network is used to extract features of small molecule compounds in the positive samples and features of small molecule compounds in the negative samples;
[0082] The second graph neural network is used to extract features of solvent molecules in the positive samples and features of solvent molecules in the negative samples;
[0083] The fully connected neural network is used to extract feature vectors that connect the features of the small molecule compound and the features of the solvent molecule, and outputs the probability of a specific small molecule compound forming a solvate with different solvent molecules based on the feature vectors.
[0084] Optionally, the probability that the specific small molecule compound will form a solvate with different solvent molecules is output in the form of a score.
[0085] Optionally, the acquisition module 61 is used to collect training data from at least one of the following sources: papers, databases, and experimental data of solvates.
[0086] like Figure 7 As shown in the embodiments of this application, a solvate prediction device 70 based on a solvate prediction model is also provided, comprising:
[0087] Module 71 is used to acquire the solvate prediction model;
[0088] The solvate prediction module 72 is used to input the small molecule compound to be predicted into the solvate prediction model and predict recommended solvent molecules that may form solvates with the small molecule compound to be predicted, based on the solvate prediction model; or, it is used to input the small molecule compound to be predicted and candidate solvent molecules into the solvate prediction model and predict the probability that the small molecule compound to be predicted and the candidate solvent molecules will form solvates based on the solvate prediction model.
[0089] This application can also be implemented as a computer-readable storage medium (or a non-transitory machine-readable storage medium or a machine-readable storage medium) storing executable code (or computer program or computer instruction code) thereon, which, when executed by a processor of an electronic device (or server, etc.), causes the processor to perform part or all of the steps of the above-described method according to this application.
[0090] The various embodiments of this application have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.
Claims
1. A method for generating a solvate prediction model, characterized in that, include: Acquire training data, which includes positive samples and negative samples. The positive samples include multiple small molecule compounds and information on solvent molecules that can form solvates with the multiple small molecule compounds. The negative samples include multiple small molecule compounds and information on solvent molecules that cannot form solvates with the multiple small molecule compounds. An initial neural network model is trained based on the training data to obtain a solvate prediction model.
2. The method according to claim 1, characterized in that, The small molecule compounds in the positive and / or negative samples are represented by SMILES; and / or, The solvates in the positive and / or negative samples are represented by SMILES.
3. The method according to claim 1, characterized in that, The initial neural network model includes a first graph neural network, a second graph neural network, and a fully connected neural network; wherein... The first graph neural network is used to extract features of small molecule compounds in the positive samples and features of small molecule compounds in the negative samples; The second graph neural network is used to extract features of solvent molecules in the positive samples and features of solvent molecules in the negative samples; The fully connected neural network is used to extract feature vectors that connect the features of the small molecule compound and the features of the solvent molecule, and outputs the probability that the small molecule compound will form solvates with different solvent molecules based on the feature vectors.
4. The method according to claim 3, characterized in that, The probability that the small molecule compound will form a solvate with different solvent molecules is output in the form of a score.
5. The method according to claim 1, characterized in that, The acquisition of training data includes: Training data can be collected from at least one of the following sources: papers, databases, and experimental data on solvates.
6. A solvate prediction method based on a solvate prediction model, characterized in that, include: A solvate prediction model is obtained; wherein the solvate prediction model is obtained by training an initial neural network model based on training data; the training data includes positive samples and negative samples, the positive samples include multiple small molecule compounds and information on solvent molecules that can form solvates with the multiple small molecule compounds respectively, and the negative samples include multiple small molecule compounds and information on solvent molecules that cannot form solvates with the multiple small molecule compounds respectively. The solvate prediction model takes a small molecule compound to be predicted as input, and predicts recommended solvent molecules that may form solvates with the small molecule compound. The solvate prediction model is input with the small molecule compound to be predicted and the candidate solvent molecule, and the probability of the small molecule compound to be predicted forming a solvate with the candidate solvent molecule is predicted by the solvate prediction model.
7. The method according to claim 6, characterized in that, The solvate prediction model includes a first graph neural network, a second graph neural network, and a fully connected neural network; wherein... When the solvate prediction model is input with a small molecule compound to be predicted and a candidate solvent molecule, the step of predicting the probability of the small molecule compound to be predicted forming a solvate with the candidate solvent molecule through the solvate prediction model includes: The features of the small molecule compound to be predicted are extracted using the first graph neural network; The features of the candidate solvent molecules are extracted using the second graph neural network; The fully connected neural network extracts feature vectors that connect the features of the small molecule compound to be predicted and the features of the candidate solvent molecule, and outputs the probability that the small molecule compound to be predicted will form a solvate with the candidate solvent molecule based on the feature vectors.
8. The method according to claim 6, characterized in that, The solvate prediction model includes a first graph neural network, a second graph neural network, and a fully connected neural network; wherein... When the solvate prediction model is input with a small molecule compound to be predicted, the step of predicting recommended solvent molecules that may form solvates with the small molecule compound through the solvate prediction model includes: The features of the small molecule compound to be predicted are extracted using the first graph neural network; The second graph neural network provides features for a variety of candidate solvent molecules; The fully connected neural network extracts feature vectors that connect the features of the small molecule compound to be predicted and the features of the multiple candidate solvent molecules, and obtains the probability that the small molecule compound to be predicted will form solvates with the multiple candidate solvent molecules based on the feature vectors. Based on the stated probability, recommended solvent molecules are obtained that may form solvates with the predicted small molecule compounds.
9. An apparatus for generating a solvate prediction model, characterized in that, The method includes a memory and a processor, wherein executable code is stored in the memory, and when the executable code is processed by the processor, the processor causes the processor to perform the method as described in any one of claims 1 to 5.
10. A solvate prediction device based on a solvate prediction model, characterized in that, The method includes a memory and a processor, wherein executable code is stored in the memory, and when the executable code is processed by the processor, the processor causes the processor to perform the method as described in any one of claims 6 to 8.
11. An apparatus for obtaining a predictive model of a solvate, characterized in that, include: An acquisition module is used to acquire training data, which includes positive samples and negative samples. The positive samples include multiple small molecule compounds and information on solvent molecules that can form solvates with the multiple small molecule compounds. The negative samples include multiple small molecule compounds and information on solvent molecules that cannot form solvates with the multiple small molecule compounds. The training module is used to train an initial neural network model based on the training data to obtain a solvate prediction model.
12. A solvate prediction device based on a solvate prediction model, characterized in that, include: An acquisition module is used to acquire a solvate prediction model; wherein the solvate prediction model is obtained by training an initial neural network model based on training data; the training data includes positive samples and negative samples, the positive samples include multiple small molecule compounds and information on solvent molecules that can form solvates with the multiple small molecule compounds respectively, and the negative samples include multiple small molecule compounds and information on solvent molecules that cannot form solvates with the multiple small molecule compounds respectively. The solvate prediction module is used to input the small molecule compound to be predicted into the solvate prediction model, and predict recommended solvent molecules that may form solvates with the small molecule compound to be predicted, based on the solvate prediction model; or, it is used to input the small molecule compound to be predicted and candidate solvent molecules into the solvate prediction model, and predict the probability that the small molecule compound to be predicted and the candidate solvent molecules will form solvates based on the solvate prediction model.
13. A computer-readable storage medium, characterized in that, The device stores executable code that, when executed by a processor of the electronic device, causes the electronic device to perform the method as described in any one of claims 1 to 8.