Method and apparatus for predicting activity of a compound on a hormone receptor

Abstract

The application provides a compound hormone receptor activity prediction method and device, the method comprises the following steps: obtaining sample compound data with known estrogen and androgen receptor activity; drawing an IC50 value distribution map based on the IC50 value in the sample compound data, and determining a demarcation point of the strength of the estrogen and androgen receptor activity according to the IC50 value distribution map; classifying the compounds in the sample compound data into strong and weak estrogen and androgen receptor activity according to the demarcation point; converting the SMILES code of the compounds in the sample compound data into corresponding fingerprint information; describing the compounds in the sample compound data with the fingerprint information, and labeling the compounds in the sample compound data with the activity strength / weakness as a label to update the sample compound data; training a machine learning model through the updated sample compound data to obtain a compound hormone receptor activity prediction model; and predicting the hormone receptor activity of a to-be-predicted compound with the prediction model.

Description

Technical Field

[0001] This invention relates to the field of machine learning technology, specifically to a method and device for predicting the activity of compound hormone receptors. Background Technology

[0002] Many natural and non-natural compounds are widely used in daily life, such as preservatives, UV filters, plasticizers, fragrances, antibacterial agents, pesticides, and flame retardants in personal care and consumer products, as well as in food. It is noteworthy that many compounds, as endocrine disrupting chemicals (EDCs), can interfere with the synthesis and action of human hormones, leading to various diseases such as decreased fertility, cancer, and even infant mortality. According to the Endocrine Society, endocrine disrupting chemicals are "exogenous, unnatural chemical substances, or mixtures of chemical substances, that interfere with any aspect of hormone action." Currently, many methods are employed both domestically and internationally to determine whether a compound has endocrine activity. Endocrine disruptor screening programs and 21st Century Toxicology Testing Projects have established various in vitro or in vivo detection methods to measure the potential effects of chemicals on the endocrine systems of humans or wild animals. Although cell culture, especially stem cell culture, can replace animal experiments, thus shortening the evaluation process and improving experimental sensitivity, experimental assessment of compound toxicity remains time-consuming and laborious. Using mathematical models to predict compound activity has become a research hotspot in computer-aided drug design. Data modeling tools can expand existing experimental data within a certain framework, thereby reducing the consumption of human and material resources.

[0003] Quantitative structure-activity relationship (QSAR) is widely used to predict the activity of compounds on endocrine proteins such as estrogen and androgen receptors. For example, projects such as the collaborative estrogen receptor activity prediction project and the co-modeling project for androgen receptor activity prediction utilize estrogen or androgen receptor activity predictions trained using different QSAR methods. This method is often combined with machine learning algorithms. Based on a review of domestic and international applications of machine learning for predicting the estrogen and androgen receptor activity of compounds, this paper establishes an activity prediction model using support vector machines, random forests, and K-nearest neighbor classifiers, based on the quantitative structure-activity relationship method and a database. This model is beneficial for high-throughput screening of compounds.

[0004] In summary, while machine learning algorithms have significantly improved the process of chemical evaluation and rapid property screening through quantitative structure-activity relationship (QSAR) prediction models, their complexity, operational difficulties, and the need for specialized computational resources and a certain level of disciplinary background limit their application. Therefore, a model that can accept richer input information is needed to achieve a direct mapping from chemical structure to properties, reduce the knowledge requirements for model use, and broaden the model's applicability. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method and apparatus for predicting the activity of hormone receptors of compounds, which can predict the activity of hormone receptors of compounds simply, quickly, accurately, and efficiently.

[0006] The technical solution adopted in this invention is as follows:

[0007] A method for predicting hormone receptor activity of compounds includes the following steps: acquiring sample compound data with known estrogen and androgen receptor activities, wherein the sample compound data includes the SMILES code and IC50 value of the compound; plotting an IC50 value distribution map based on the IC50 values ​​in the sample compound data, and determining the dividing point between strong and weak estrogen and androgen receptor activities according to the IC50 value distribution map; classifying the compounds in the sample compound data into those with strong androgen receptor activities and those with weak estrogen and androgen receptor activities according to the dividing point; converting the SMILES codes of the compounds in the sample compound data into corresponding fingerprint information; describing the compounds in the sample compound data with the fingerprint information, and labeling the compounds in the sample compound data with labels of strong androgen receptor activities and weak estrogen and androgen receptor activities to update the sample compound data; training a machine learning model with the updated sample compound data to obtain a compound hormone receptor activity prediction model; converting the compound to be predicted into a SMILES code and then into corresponding fingerprint information; inputting the fingerprint information corresponding to the compound to be predicted into the compound hormone receptor activity prediction model to obtain the prediction result of strong / weak estrogen and androgen receptor activities of the compound to be predicted.

[0008] The sample compound data was obtained from the BindingDB database.

[0009] The distribution of the IC50 values ​​was plotted using Pandas.

[0010] The fingerprint information is either an ECFP fingerprint or a MACCS fingerprint.

[0011] The machine learning model is a KNN model, a RF model, or an SVM model.

[0012] A compound hormone receptor activity prediction device includes: an acquisition module for acquiring sample compound data with known estrogen and androgen receptor activities, wherein the sample compound data includes the compound's SMILES code and IC50 value; a plotting module for plotting an IC50 value distribution map based on the IC50 values ​​in the sample compound data, and determining the dividing point between strong and weak estrogen and androgen receptor activities based on the IC50 value distribution map; a classification module for classifying the compounds in the sample compound data into those with strong androgen receptor activities and those with weak estrogen and androgen receptor activities based on the dividing point; a first conversion module for converting the SMILES codes of the compounds in the sample compound data into corresponding fingerprint information; and an update module. The system comprises the following modules: a module for describing compounds in the sample compound data using the fingerprint information, and labeling compounds in the sample compound data with labels indicating strong and weak estrogen receptor activity, thereby updating the sample compound data; a training module for training a machine learning model using the updated sample compound data to obtain a compound hormone receptor activity prediction model; a second conversion module for converting the compound to be predicted into SMILES codes and then into corresponding fingerprint information; and a prediction module for inputting the fingerprint information corresponding to the compound to be predicted into the compound hormone receptor activity prediction model to obtain a prediction result indicating strong / weak estrogen receptor activity of the compound to be predicted.

[0013] The acquisition module retrieves the sample compound data from the BindingDB database.

[0014] The plotting module uses Pandas to draw the IC50 value distribution map.

[0015] The fingerprint information is either an ECFP fingerprint or a MACCS fingerprint.

[0016] The machine learning model is a KNN model, a RF model, or an SVM model.

[0017] The beneficial effects of this invention are:

[0018] The compound hormone receptor activity prediction method and apparatus of this invention, through SMILES encoding and fingerprint information conversion, employs a machine learning model that eliminates the need for manually defined quantifiable structural parameters as molecular descriptors, saving time and computational resources for molecular descriptor calculation and selection, and requiring less computational chemistry background for application. By plotting IC50 value distribution maps and classifying activity strength at defined cutoff points, a direct mapping between chemical structure and estrogen / androgen activity can be established. This allows for reverse inference from chemical activity to chemical structure, identifying the chemical structural features of estrogen / androgen activity, which is helpful for the design and discovery of compounds with specific activity. It is suitable for large-scale screening of estrogen activity in chemicals, is simple and fast, and has high accuracy and efficiency. It has broad application prospects in compound risk assessment, environmental safety assessment, and other fields. Attached Figure Description

[0019] Figure 1 This is a flowchart of the compound hormone receptor activity prediction method according to an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of a compound fingerprint according to an embodiment of the present invention;

[0021] Figure 3 This is a Roc curve diagram of a specific embodiment of the present invention;

[0022] Figure 4 This is a block diagram of a compound hormone receptor activity prediction device according to an embodiment of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] like Figure 1 As shown, the compound hormone receptor activity prediction method of this invention includes the following steps:

[0025] S1, Obtain sample compound data with known estrogen and androgen receptor activities, wherein the sample compound data includes the compound's SMILES code and IC50 value.

[0026] In one embodiment of the present invention, sample compound data can be obtained from the BindingDB database. Furthermore, the compounds in the sample compound data can be converted into SMILES codes, and the IC50 values ​​can be logarithmically taken for subsequent data processing operations.

[0027] S2. Based on the IC50 values ​​in the sample compound data, plot the IC50 value distribution map, and determine the dividing point between the strength of estrogen and androgen receptor activity based on the IC50 value distribution map.

[0028] S3, based on the dividing point, classify the compounds in the sample compound data into those with strong androgen receptor activity and those with weak androgen receptor activity.

[0029] In one embodiment of the present invention, an IC50 value distribution map can be plotted using Pandas. After obtaining the IC50 value distribution map, the median data in the distribution map can be used as a dividing point. Compounds exceeding the dividing point are classified as having strong estrogen receptor activity, while compounds not exceeding the dividing point are classified as having weak estrogen receptor activity.

[0030] S4 converts the SMILES codes of compounds in the sample compound data into corresponding fingerprint information.

[0031] In one embodiment of the present invention, the fingerprint information may be an ECFP fingerprint or a MACCS fingerprint. For example... Figure 2 As shown, fingerprint information can represent the basic structural characteristics of a compound.

[0032] Specifically, the SMILES codes of compounds in the sample compound data can be converted into ECFP fingerprints using the Morgan algorithm in the RDKit tool. Alternatively, the SMILES codes of compounds in the sample compound data can be converted into MACCS fingerprints using the DeepChem open-source software package developed by Pande.

[0033] S5 describes the compounds in the sample compound data using fingerprint information, and labels the compounds in the sample compound data with strong and weak estrogen receptor activity and estrogen receptor activity as tags to update the sample compound data.

[0034] S6. The machine learning model is trained using the updated sample compound data to obtain a compound hormone receptor activity prediction model.

[0035] The updated sample compound data can be divided into a training set and a validation set in an 8:2 ratio. The machine learning model is trained using the training set, and the training effect is verified using the validation set.

[0036] In one embodiment of the present invention, the machine learning model may be a KNN model, a RF model, or an SVM model. Users can select the appropriate machine learning model based on the characteristics of the dataset.

[0037] Furthermore, in one embodiment of the present invention, the prediction results of the compound hormone receptor activity prediction model can be evaluated using the Sklearn.metric.classification_report tool, and the corresponding RoC curve can be output using pyPlot. A specific embodiment of the RoC curve of the present invention is shown below. Figure 3 As shown in Table 1, the evaluation indicators for a certain model are as follows.

[0038] Table 1

[0039]

[0040] S7 converts the compound to be predicted into SMILES codes and then into corresponding fingerprint information.

[0041] S8. Input the fingerprint information corresponding to the compound to be predicted into the compound hormone receptor activity prediction model to obtain the prediction result of the strong / weak estrogen receptor activity of the compound to be predicted.

[0042] The compound hormone receptor activity prediction method according to embodiments of the present invention, through SMILES encoding and fingerprint information conversion, employs a machine learning model that eliminates the need for manually defined quantifiable structural parameters as molecular descriptors, saving time and computational resources for molecular descriptor calculation and selection, and requiring less computational chemistry background for application. By plotting IC50 value distribution maps and classifying activity strength at defined boundary points, a direct mapping between chemical structure and estrogen / androgen activity can be established. This allows for reverse inference from chemical activity to chemical structure, identifying chemical structural features of estrogen / androgen activity, which is helpful for the design and discovery of compounds with specific activity. It is suitable for large-scale screening of estrogen activity in chemicals, is simple and fast, and has high accuracy and efficiency. It has broad application prospects in compound risk assessment, environmental safety assessment, and other fields.

[0043] Corresponding to the compound hormone receptor activity prediction method in the above embodiments, the present invention also proposes a compound hormone receptor activity prediction device.

[0044] like Figure 4As shown, the compound hormone receptor activity prediction device of this invention includes: an acquisition module 10, a plotting module 20, a classification module 30, a first conversion module 40, an update module 50, a training module 60, a second conversion module 70, and a prediction module 80. The acquisition module 10 is used to acquire sample compound data with known estrogen and androgen receptor activities, wherein the sample compound data includes the compound's SMILES code and IC50 value; the plotting module 20 is used to plot an IC50 value distribution map based on the IC50 values ​​in the sample compound data, and determine the dividing point between strong and weak estrogen and androgen receptor activities based on the IC50 value distribution map; the classification module 30 is used to classify the compounds in the sample compound data into those with strong androgen receptor activities and those with weak estrogen and androgen receptor activities based on the dividing point; the first conversion module 40 is used to convert the SMILES codes of the compounds in the sample compound data into corresponding fingerprint information; the update module 50 is used to... The fingerprint information describes the compounds in the sample compound data, and labels the compounds in the sample compound data with strong and weak estrogen receptor activity as labels to update the sample compound data; the training module 60 is used to train the machine learning model with the updated sample compound data to obtain the compound hormone receptor activity prediction model; the second conversion module 70 is used to convert the compound to be predicted into SMILES encoding and into corresponding fingerprint information; the prediction module 80 is used to input the fingerprint information corresponding to the compound to be predicted into the compound hormone receptor activity prediction model to obtain the prediction result of strong / weak estrogen receptor activity of the compound to be predicted.

[0045] In one embodiment of the present invention, the acquisition module 10 can acquire sample compound data from the BindingDB database. Furthermore, the acquisition module 10 can convert the compounds in the sample compound data into SMILES codes, and can take the logarithmic value of the IC50 value for subsequent data processing operations.

[0046] In one embodiment of the present invention, the plotting module 20 can plot an IC50 value distribution map using Pandas. After obtaining the IC50 value distribution map, the median data in the distribution map can be used as a dividing point. Compounds exceeding the dividing point are classified as having strong estrogen receptor activity, while compounds not exceeding the dividing point are classified as having weak estrogen receptor activity.

[0047] In one embodiment of the present invention, the fingerprint information may be an ECFP fingerprint or a MACCS fingerprint. For example... Figure 2 As shown, fingerprint information can represent the basic structural characteristics of a compound.

[0048] Specifically, the first conversion module 40 can convert the SMILES codes of compounds in the sample compound data into ECFP fingerprints using the Morgan algorithm in the RDKit tool. Alternatively, the first conversion module 40 can convert the SMILES codes of compounds in the sample compound data into MACCS fingerprints using the DeepChem open-source software package developed by Pande.

[0049] The updated sample compound data can be divided into a training set and a validation set in an 8:2 ratio. The machine learning model is trained using the training set, and the training effect is verified using the validation set.

[0050] In one embodiment of the present invention, the machine learning model may be a KNN model, a RF model, or an SVM model. Users can select the appropriate machine learning model based on the characteristics of the dataset.

[0051] Furthermore, in one embodiment of the present invention, the prediction results of the compound hormone receptor activity prediction model can be evaluated using the Sklearn.metric.classification_report tool, and the corresponding RoC curve can be output using pyPlot. A specific embodiment of the RoC curve of the present invention is shown below. Figure 3 As shown in Table 1, the evaluation indicators for a certain model are as follows.

[0052] The compound hormone receptor activity prediction device according to embodiments of the present invention, through the conversion of SMILES encoding and fingerprint information, employs a machine learning model that does not require manually defined quantifiable structural parameters as molecular descriptors, saving time and computational resources for molecular descriptor calculation and selection, and requiring less computational chemistry background for application; by plotting IC50 value distribution maps and classifying activity strength at defined boundary points, it can directly establish a mapping between chemical structure and estrogen and androgen activity, enabling reverse inference from chemical activity to chemical structure, finding chemical structural features of estrogen and androgen activity, which is helpful for the design and discovery of specific active compounds; it is suitable for large-scale screening of chemical estrogen activity, is simple and fast, and has high accuracy and efficiency; it has broad application prospects in compound risk assessment, environmental safety assessment and other fields.

[0053] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. "A plurality of" means two or more, unless otherwise explicitly specified.

[0054] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0055] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can mean that the first feature is in direct contact with the second feature, or that the first feature is in indirect contact with the second feature through an intermediate medium. Furthermore, "above," "over," and "on top" of the second feature can mean that the first feature is directly above or diagonally above the second feature, or simply that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature can mean that the first feature is directly below or diagonally below the second feature, or simply that the first feature is at a lower horizontal level than the second feature.

[0056] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0057] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0058] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0059] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0060] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0061] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0062] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method of predicting compound hormone receptor activity, characterized by, Includes the following steps: Acquire sample compound data with known estrogen and androgen receptor activities, wherein the sample compound data includes the compound's SMILES code and IC50 value; Based on the IC50 values ​​in the sample compound data, an IC50 value distribution map was plotted, and the dividing point between the strength of estrogen and androgen receptor activity was determined according to the IC50 value distribution map. Based on the dividing point, the compounds in the sample compound data are classified into those with strong androgen receptor activity and those with weak androgen receptor activity. The SMILES codes of the compounds in the sample compound data are converted into corresponding fingerprint information; The fingerprint information is used to describe the compounds in the sample compound data, and the compounds in the sample compound data are labeled with strong and weak estrogen receptor activity and estrogen receptor activity, so as to update the sample compound data. The machine learning model was trained using the updated sample compound data to obtain a compound hormone receptor activity prediction model. The compound to be predicted is converted into SMILES codes and then into corresponding fingerprint information; The fingerprint information corresponding to the compound to be predicted is input into the hormone receptor activity prediction model of the compound to obtain the prediction result of the strong / weak estrogen receptor activity of the compound to be predicted.

2. The method for predicting hormone receptor activity of compounds according to claim 1, characterized in that, The sample compound data was obtained from the BindingDB database.

3. The method for predicting hormone receptor activity of compounds according to claim 1, characterized in that, The distribution of the IC50 values ​​was plotted using Pandas.

4. The method for predicting hormone receptor activity of compounds according to claim 1, characterized in that, The fingerprint information is either an ECFP fingerprint or a MACCS fingerprint.

5. The method for predicting hormone receptor activity of compounds according to claim 1, characterized in that, The machine learning model is a KNN model, a RF model, or an SVM model.

6. A compound hormone receptor activity prediction device, characterized in that, include: An acquisition module is used to acquire sample compound data with known estrogen and androgen receptor activities, wherein the sample compound data includes the compound's SMILES code and IC50 value; The plotting module is used to plot an IC50 value distribution map based on the IC50 values ​​in the sample compound data, and to determine the dividing point between the strength of estrogen and androgen receptor activity based on the IC50 value distribution map. A classification module is used to classify compounds in the sample compound data into those with strong androgen receptor activity and those with weak androgen receptor activity based on the dividing point. The first conversion module is used to convert the SMILES codes of compounds in the sample compound data into corresponding fingerprint information; An update module is used to describe the compounds in the sample compound data with the fingerprint information, and to label the compounds in the sample compound data with strong and weak estrogen receptor activity and estrogen receptor activity as tags, so as to update the sample compound data. The training module is used to train the machine learning model with updated sample compound data to obtain a compound hormone receptor activity prediction model. The second conversion module is used to convert the compound to be predicted into SMILES codes and then into corresponding fingerprint information; The prediction module is used to input the fingerprint information corresponding to the compound to be predicted into the hormone receptor activity prediction model of the compound to obtain the prediction result of the strong / weak estrogen receptor activity of the compound to be predicted.

7. The compound hormone receptor activity prediction device according to claim 6, characterized in that, The acquisition module retrieves the sample compound data from the BindingDB database.

8. The compound hormone receptor activity prediction device according to claim 6, characterized in that, The plotting module uses Pandas to draw the IC50 value distribution map.

9. The compound hormone receptor activity prediction device according to claim 6, characterized in that, The fingerprint information is either an ECFP fingerprint or a MACCS fingerprint.

10. The compound hormone receptor activity prediction device according to claim 6, characterized in that, The machine learning model is a KNN model, a RF model, or an SVM model.

Images