Liposome construction recommendation method and device based on drug molecular structure

Abstract

The application discloses a liposome construction recommendation method and device based on drug molecular structure, and the recommendation method is characterized in that, after receiving the molecular information of a target drug input by a user, firstly, the molecular information is subjected to data preprocessing; then, a pre-trained liposome recommendation model is used to make a recommendation based on standard molecular characteristic data; finally, the construction parameters are optimized based on the target liposome construction type, so that the target liposome construction parameters are obtained. The pre-trained neural network model is used to make a recommendation on the construction type of the liposome based on the drug molecular information, and the construction parameters of the liposome are optimized to obtain the optimal liposome construction parameters. The drug molecular structure characteristics are directly associated with the liposome construction scheme, and the efficiency and accuracy of the liposome construction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, more particularly, to a liposome construction recommendation method and device based on drug molecular structure. BACKGROUND

[0002] As a new type of drug carrier, liposomes are widely used in drug delivery and controlled release due to their good biocompatibility, drug protection and targeting. The key to liposome construction is to determine the appropriate construction type and key parameters according to the structure and physicochemical properties of different drug molecules to achieve the best drug delivery effect.

[0003] The construction of liposomes involves complex processes and diverse parameter adjustments. Currently, liposome construction mainly relies on expert experience to design liposome components, ratios and process parameters during construction. However, relying on expert experience or literature queries for liposome construction, the extraction and analysis of drug molecule characteristics lack standardization, making it difficult to directly associate drug molecule characteristics with appropriate liposome construction schemes, resulting in low construction efficiency. SUMMARY

[0004] Therefore, the present application provides a liposome construction recommendation method and device based on drug molecular structure and physicochemical properties to solve the problem of low construction efficiency caused by the difficulty of directly associating drug molecule characteristics with appropriate liposome construction schemes in the existing liposome construction method.

[0005] To achieve the above-mentioned purpose, the present scheme is as follows:

[0006] A liposome construction recommendation method based on drug molecular structure, the method comprising:

[0007] Receiving the molecular information of the target drug input by the user;

[0008] Data preprocessing of the molecular information to obtain standard molecular feature data;

[0009] Recommending based on the standard molecular feature data through the pre-trained liposome recommendation model to obtain the target liposome construction type;

[0010] Optimizing the construction parameters based on the target liposome construction type to obtain the target liposome construction parameters.

[0011] Preferably, the process of data preprocessing of the molecular information comprises:

[0012] Converting the molecular information to a standard format;

[0013] Removing interference information in the molecular information;

[0014] molecular energy optimization is performed on the target drug molecule;

[0015] physicochemical characteristics of the target drug molecule are extracted to obtain standard molecular feature data.

[0016] Preferably, the standard molecular feature data includes at least one of molecular weight, LogP, polar surface area, number of hydrogen bond donors, number of hydrogen bond acceptors, charge distribution, and free energy.

[0017] Preferably, the method further comprises:

[0018] The molecular information and the target liposome construction parameters are stored.

[0019] Preferably, the target liposome construction type is a single-layer liposome, a multi-layer liposome, a targeted liposome, a temperature-sensitive liposome, a PEG liposome, or a dual-liposome.

[0020] Preferably, the construction parameters include particle size, lipid / cholesterol ratio, and charge potential.

[0021] Preferably, the method further comprises:

[0022] The target liposome construction type and the target liposome construction parameters are displayed.

[0023] A liposome construction recommendation device based on drug molecule structure, the device comprising:

[0024] A drug molecule input module for receiving user input of molecular information of a target drug;

[0025] A drug molecule preprocessing module for data preprocessing of the molecular information to obtain standard molecular feature data;

[0026] A liposome construction recommendation module for recommendation based on the standard molecular feature data by a pre-trained liposome recommendation model to obtain a target liposome construction type;

[0027] A parameter optimization module for optimizing construction parameters based on the target liposome construction type to obtain target liposome construction parameters.

[0028] Preferably, the device further comprises:

[0029] A data storage and management module for storing the molecular information and the target liposome construction parameters.

[0030] Preferably, the device further comprises:

[0031] A result display module for displaying the target liposome construction type and the target liposome construction parameters.

[0032] According to the specific embodiments of the present application, the following technical effects are disclosed:

[0033] The liposome construction recommendation method based on the molecular structure of a drug provided by the present application comprises the following steps: first, receiving the molecular information of a target drug input by a user; then, performing data preprocessing on the molecular information; then, recommending based on standard molecular feature data through a pre-trained liposome recommendation model; and finally, optimizing the construction parameters based on the target liposome construction type to obtain the target liposome construction parameters. The neural network model is pre-trained, the molecular information of the drug is used to recommend the structure type of the constructed liposome, the construction parameters of the liposome are optimized, and the optimal liposome construction parameters are obtained. The molecular structure characteristics and physicochemical properties of the drug are directly associated with the liposome construction scheme, and the efficiency and accuracy of the liposome construction are improved. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of the provided drawings.

[0035] Figure 1 A flow chart of a liposome construction recommendation method based on the molecular structure of a drug is provided for the embodiments of the present application.

[0036] Figure 2 A structure schematic diagram of a liposome construction recommendation device based on the molecular structure of a drug is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0037] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0038] First, the liposome construction recommendation method based on the molecular structure of a drug provided by the present application will be introduced as follows. Figure 1 The liposome construction recommendation method based on the molecular structure of a drug provided by the present application will be introduced as follows. Figure 1 As shown in the figure, the recommendation method comprises the following steps:

[0039] Step S01, receiving the molecular information of a target drug input by a user.

[0040] Specifically, a user can input a file containing molecular information of a target drug in different formats. For example, SMILES, PDB, MOL format, or a molecular structure drawn by chemdraw.

[0041] Step S02, data preprocessing is performed on the molecular information.

[0042] Specifically, first, format conversion is performed to convert the molecular information into a standard format. The molecular structure format conversion can be performed by Open Babel to convert the format of the input file containing the molecular information (SMILES, PDB, MOL format) into a unified standard format.

[0043] Then, interference information in the molecular information is removed. The interference information in the molecular information can be removed by PyMOL, which includes water molecules, organic solvents, unnecessary small molecules, etc. Thus, the purity of the drug molecular structure is ensured.

[0044] Then, the molecular energy of the target drug is optimized. The energy optimization can be performed by ORCA, and a relatively stable molecular structure of the target drug can be obtained. The process provides basic data for subsequent calculations and experiments. The entire process can be visualized by PyMOL.

[0045] Finally, the physicochemical characteristics of the target drug molecule are extracted. The physicochemical characteristics of the target drug molecule can be extracted by RDKit to obtain standard molecular characteristic data. For example, at least one of the following: molecular weight, LogP, polar surface area, number of hydrogen bond donors, number of hydrogen bond acceptors, charge distribution, and free energy.

[0046] Molecular weight (Molecular Weight, MW): basic information of the size of the target drug molecule.

[0047] LogP: the liposolubility of the target drug molecule, which mainly affects the behavior of the target drug molecule in the liposome.

[0048] Polar surface area (Polar Surface Area, PSA): used to predict the membrane penetration ability of the target drug molecule.

[0049] Number of hydrogen bond donors (Hydrogen Bond Donors, HBD): interaction characteristics of the target drug molecule with the surrounding environment.

[0050] Number of hydrogen bond acceptors (Hydrogen Bond Acceptors, HBA): the ability of the target drug molecule to interact with solvents and lipids.

[0051] Charge Distribution: The surface charge of a drug molecule of interest influences its behavior in aqueous solution.

[0052] Free Energy: An important parameter that describes the stability of a drug molecule of interest.

[0053] Step S03, the standard molecular feature data is used to make recommendations based on the pre-trained liposome recommendation model.

[0054] Specifically, the pre-trained liposome recommendation model is used to make recommendations for liposome construction types based on standard molecular feature data, resulting in a target liposome construction type. The liposome recommendation model can be a feedforward neural network model built using TensorFlow. The liposome recommendation model is trained based on the structural feature data of drug molecules, with the structural feature data of drug molecules as training samples and the liposome construction type as training labels. The liposome construction type includes single-layer liposomes, multi-layer liposomes, targeted liposomes, temperature-sensitive liposomes, PEG liposomes, and double-loaded liposomes, among others, and the target liposome construction type is one of them.

[0055] Step S04, the construction parameters are optimized based on the target liposome construction type.

[0056] Specifically, the pre-trained parameter optimization model can be used to optimize the construction parameters based on spatial coordinate information, target liposome construction type, and standard molecular feature data, etc., to obtain target liposome construction parameters. Scikit-learn can be used to optimize the construction parameters based on spatial coordinate information, target liposome construction type, and standard molecular feature data, etc., to ensure the stability and high encapsulation efficiency of liposome construction. The construction parameters include particle size, lipid / cholesterol ratio, and charge potential, etc.

[0057] Particle Size: The physical size of a liposome (unit: nm).

[0058] Lipid-to-Cholesterol Ratio: The common range is 1:1 to 5:1.

[0059] Zeta Potential: The surface charge value of a liposome, usually -50 to 100 mV.

[0060] The liposome construction recommendation method based on the drug molecular structure provided by the embodiment of the present application can directly associate the drug molecular structure characteristics with the liposome construction scheme, improve the efficiency and accuracy of the liposome construction.

[0061] In order to more accurately perform the liposome construction recommendation, on the basis of the liposome construction recommendation method based on the drug molecular structure of the foregoing embodiment, the liposome construction recommendation method of the embodiment of the present application can further include the following steps:

[0062] The molecular information and the target liposome construction parameter are stored.

[0063] Specifically, the molecular information and the target liposome construction parameter can be stored through a pre-constructed database. The drug molecular information and the liposome construction recommendation data can be stored through a MongoDB database. The user-uploaded molecular information, the extracted standard molecular feature data, the recommended target liposome construction type and the target liposome construction parameter can be used for subsequent model training and recommendation improvement.

[0064] The database can contain the molecular information of the related drugs from the known targeted drug sources collected from chemdraw. The information collected by the database mainly includes two parts. One is the related basic information of the collected drug molecules, for example, the name, spatial coordinates, related id and the like of the drug molecules, which are used as information annotation and as information carriers. The other is the extracted standard molecular feature data and the like.

[0065] Further, the feedback data of the user on the target liposome construction type and the target liposome construction parameter can also be received and stored. The feedback data can be used to improve the liposome recommendation model and the parameter optimization model, and improve the accuracy of the liposome construction scheme recommendation.

[0066] In addition, the following steps can also be included:

[0067] The target liposome construction type and the target liposome construction parameter are displayed.

[0068] Specifically, the recommended liposome construction protocols are visualized. This visualization can be implemented using the React Native framework. The displayed content can include: the target liposome construction type, the target liposome construction parameters, and the recommended range of construction parameters.

[0069] Next, this invention will introduce a liposome recommendation model. This model can be constructed using TensorFlow, a classic binary classification model used for classification problems. In this invention, liposome construction types are recommended based on input feature data.

[0070] The input feature data can be the spatial coordinates of drug molecules and standard molecular feature data. The number of output layer nodes corresponds to the number of liposome construction types. Softmax is used as the activation function to convert the value of each output node into a probability value, the sum of which is 1. The output value of each node represents the prediction confidence for that construction type. Categorical Crossentropy can be used as the loss function to train the liposome recommendation model. The formula for the liposome recommendation model is as follows:

[0071]

[0072] in, z i It is the first i The original values ​​(logits) of each output node. z j No. j The original values ​​of the output nodes, where n is the number of output nodes. P ( y i ) is the output node i The predicted probability of the corresponding liposome construction type.

[0073] Liposome recommendation model training data processing procedure:

[0074] Deduplication: Check if there are duplicate entries in the dataset and remove duplicate data.

[0075] Missing value handling: If the physicochemical properties of some molecules are missing, they can be filled using the mean, median, or interpolation method; if important fields (such as molecular weight, LogP) are missing, the incomplete data are directly removed.

[0076] Feature vector processing: convert molecular physicochemical properties into feature vectors. Among them, molecular weight, LogP, PSA, HBD, HBA, charge distribution and free energy can be directly retained according to the numerical value; charge distribution and free energy need to be normalized. If the molecular information contains chemical structure, the molecular fingerprint can be extracted as supplementary features using tools such as RDKit.

[0077] Liposome type labeling: add a liposome construction type to each drug molecule as a training label.

[0078] Data standardization: standardize all feature vectors, such as Z-Score standardization.

[0079] The parameter optimization model of the embodiment of the present application is introduced, and the parameter optimization model of the embodiment of the present application can be constructed based on Scikit-learn. The Scikit-learn model is mainly used for optimizing the construction parameters in the liposome construction process. In the present application, the spatial coordinate information, the standard molecular feature data and the target liposome construction type are input to optimize the liposome construction parameters.

[0080] The parameter optimization model can use a support vector regression model. The core of the support vector regression model is to find the best hyperplane to minimize the deviation between the predicted value and the true value. The polynomial kernel (Polynomial Kernel) is suitable for the case where there is a polynomial relationship between the input features and the target. The parameter optimization model is provided with a penalty coefficient, an error range without penalty and a kernel function parameter. Among them, the penalty coefficient is used to control the tolerance of the parameter optimization model to errors; the error range without penalty is the error range that the target value can fall into; the kernel function parameter is used to control the mapping complexity of the data.

[0081] In addition, the parameter optimization model can be evaluated by mean square error and determination coefficient. Among them, the mean square error (MSE) can be used to measure the error between the predicted value and the true value; the determination coefficient (R² Score) can be used to measure the explanatory power of the parameter optimization model to the target variable, the range is [0,1], the closer to 1 indicates that the fitting effect of the parameter optimization model is better.

[0082] Taking the number of output nodes of the liposome recommendation model as 3, corresponding to single-layer liposome, multi-layer liposome and targeted liposome, as an example, the parameter optimization model training data processing process is introduced, and the process is as follows:

[0083] The target liposome construction type output by the liposome recommendation model is converted into a category label. The corresponding relationship between each liposome construction type and the category label is as follows: single-layer liposome: 0; multi-layer liposome: 1; targeted liposome: 2.

[0084] Obtain the construction parameters of each liposome type from experiments or literature as training labels, wherein the particle size, charge potential and other parameters are normalized.

[0085] The drug molecule structure-based liposome construction recommendation device provided in the embodiments of the present application is described below, and the drug molecule structure-based liposome construction recommendation device described below can be referred to in correspondence with the drug molecule structure-based liposome construction recommendation method described above.

[0086] First, the drug molecule structure-based liposome construction recommendation method provided in the embodiments of the present application is described below. Figure 2 The drug molecule structure-based liposome construction recommendation device is introduced, as shown in Figure 2 The drug molecule structure-based liposome construction recommendation device can include:

[0087] The drug molecule input module 100 is configured to receive the molecular information of the target drug input by the user and transmit the molecular information to the drug molecule preprocessing module 200.

[0088] The drug molecule preprocessing module 200 is configured to perform data preprocessing on the molecular information to obtain standard molecular feature data.

[0089] The liposome construction recommendation module 300 is configured to perform recommendation based on the standard molecular feature data by using the pre-trained liposome recommendation model to obtain a target liposome construction type.

[0090] The parameter optimization module 400 is configured to optimize the construction parameters based on the target liposome construction type to obtain target liposome construction parameters.

[0091] Further, on the basis of the liposome construction recommendation device in the foregoing embodiments, the device can further include:

[0092] The data storage and management module is configured to store the molecular information and the target liposome construction parameters.

[0093] Further, on the basis of the liposome construction recommendation device in the foregoing embodiments, the device can further include:

[0094] The result display module is configured to display the target liposome construction type and the target liposome construction parameters.

[0095] The embodiments of the present application also provide a storage medium, which can store a program suitable for execution by a processor, and the program is used to implement each processing flow in the foregoing drug molecule structure-based liposome construction recommendation scheme.

[0096] Finally, it should be noted that the terms "first", "second", and the like, herein do not denote any order, quantity, combination, or importance, but rather are used to distinguish one element from another, and are not intended to denote the presence of any such actual relationship or order. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the presence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0097] The various embodiments in the specification are described in progressive order with reference to each embodiment, each embodiment highlighting differences from other embodiments, and the same or similar parts between the various embodiments are referred to each other.

[0098] The above description of disclosed embodiments provides enabling disclosure sufficient for one of ordinary skill in the art to practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A liposome construction recommendation method based on a drug molecule structure, characterized by, The method comprises: receiving user inputted molecular information of a target drug; data preprocessing of the molecular information to obtain standard molecular feature data; recommendation based on the standard molecular feature data by a pre-trained liposome recommendation model to obtain a target liposome construction type; the target liposome construction type is a single-layer liposome, a multi-layer liposome, a targeted liposome, a temperature-sensitive liposome, a PEG liposome or a double-loaded liposome, wherein the liposome recommendation model is a feedforward neural network model constructed by TensorFlow; the liposome recommendation model learns and predicts based on the structural feature data of the drug molecule, and is trained by taking the structural feature of the drug molecule as the training sample and taking the liposome construction type as the training label; optimizing the construction parameters based on the target liposome construction type to obtain target liposome construction parameters; optimizing the construction parameters based on the spatial coordinate information, the target liposome construction type and the standard molecular feature data by using Scikit-learn; the process of data preprocessing of the molecular information comprises: format conversion to convert the molecular information into a standard format, wherein the molecular structure format conversion is performed by Open Babel; removing interference information in the molecular information; molecular energy optimization of the target drug molecule, wherein the energy optimization is performed by ORCA; extracting physicochemical characteristics of the target drug molecule to obtain standard molecular feature data, wherein the physicochemical characteristics of the target drug molecule are extracted by RDKit; the standard molecular feature data comprises at least one of molecular weight, LogP, polar surface area, number of hydrogen bond donors, number of hydrogen bond acceptors, charge distribution and free energy; the construction parameters comprise particle size, lipid / cholesterol ratio and charge potential; the parameter optimization model is provided with a penalty coefficient, an error range without penalty and a kernel function parameter, wherein the penalty coefficient is used to control the tolerance of the parameter optimization model to error; the error range without penalty is the error range in which the target value can fall; the kernel function parameter is used to control the mapping complexity of data; and the parameter optimization model is evaluated by mean square error and determination coefficient.

2. The liposome construction recommendation method based on a drug molecule structure according to claim 1, characterized by, The method further comprises: storing the molecular information and the target liposome construction parameters.

3. The liposome construction recommendation method based on drug molecule structure according to any one of claims 1-2, characterized in that, The method further comprises: displaying the target liposome construction type and the target liposome construction parameters.

4. A liposome construction recommendation device based on a drug molecular structure, characterized by, The device comprises: a drug molecule input module for receiving user inputted molecular information of a target drug; a drug molecule preprocessing module for data preprocessing of the molecular information to obtain standard molecular feature data; a liposome construction recommendation module for recommendation based on the standard molecular feature data by a pre-trained liposome recommendation model to obtain a target liposome construction type; a parameter optimization module for optimizing the construction parameters based on the target liposome construction type to obtain target liposome construction parameters; the process of data preprocessing of the molecular information comprises: format conversion to convert the molecular information into a standard format; removing interference information in the molecular information; molecular energy optimization of the target drug molecule; extracting physicochemical characteristics of the target drug molecule to obtain standard molecular feature data; The standard molecular feature data includes at least one of molecular weight, LogP, polar surface area, number of hydrogen bond donors, number of hydrogen bond acceptors, charge distribution and free energy; The construction parameters include particle size, lipid / cholesterol ratio and charge potential; The target liposome construction type is a single-layer liposome, a multi-layer liposome, a targeted liposome, a temperature-sensitive liposome, a PEG liposome or a double-loaded liposome, wherein the liposome recommendation model is a feedforward neural network model constructed by using TensorFlow; the liposome recommendation model is trained based on the structural feature data of the drug molecules, and the structural feature data of the drug molecules is used as the training sample and the liposome construction type is used as the training label to train the liposome recommendation model; The Scikit-learn is used to optimize the construction parameters based on the spatial coordinate information, the target liposome construction type and the standard molecular feature data; The parameter optimization model is provided with a penalty coefficient, an error range without penalty and a kernel function parameter, wherein the penalty coefficient is used to control the tolerance of the parameter optimization model to errors; the error range without penalty is an error range in which the target value can fall; the kernel function parameter is used to control the mapping complexity of data; and the parameter optimization model is evaluated by using the mean square error and the determination coefficient.

5. The liposome construction recommendation apparatus based on a drug molecule structure according to claim 4, wherein, The device further comprises: A data storage and management module for storing molecular information and target liposome construction parameters.

6. The liposome construction recommendation apparatus based on drug molecule structure according to any one of claims 4-5, characterized in that, The device further comprises: A result display module for displaying the target liposome construction type and the target liposome construction parameters.

Images