Bio-enzyme catalytic function key feature screening method and device, electronic equipment and storage medium

Abstract

The invention provides a biological enzyme catalytic function key feature screening method and device, electronic equipment and a storage medium. Obtaining biological enzyme characteristics and enzyme catalytic activity; performing feature selection by adopting a recursive feature elimination method based on a random forest algorithm, and screening out key features of the biological enzyme; training a random forest model by using the screened biological enzyme key features, and obtaining a random forest model based on an optimal hyper-parameter value; and inputting the key features of the biological enzyme of the sample to be detected into the random forest model based on the optimal hyper-parameter value, obtaining a predicted value of the catalytic activity of the biological enzyme of the sample, and quantitatively calculating the contribution value of each key feature to the catalytic activity based on the predicted value. According to the method, key features influencing the catalytic function of the biological enzyme can be screened out from hundreds of millions of biological enzyme structures, charges and other features, the quantitative relation between the key features of the biological enzyme and the catalytic function is disclosed by utilizing a Shapliadditive interpretation algorithm, and guidance is provided for designing industrial biological enzymes with high catalytic activity and high selectivity.

Description

Technical Field

[0001] The present invention relates to the field of bioenzyme technology, and in particular to a method, device, electronic device and storage medium for screening key characteristics of bioenzyme catalytic function. Background Art

[0002] The information disclosed in this background technology section is only intended to enhance the understanding of the overall background of the invention, and should not be necessarily regarded as an admission or any form of suggestion that the information constitutes the prior art already known to those skilled in the art.

[0003] Enzymes are environmentally friendly catalysts with selective catalytic activity produced by living cells. Most enzymes are proteins, and a small number are DNA. The structure, charge, and electric field characteristics of enzymes are considered to be important factors affecting the catalytic function of enzymes. Therefore, revealing the relationship between the key characteristics of enzymes and their catalytic function is of great significance for the rational transformation and de novo design of enzymes.

[0004] At present, linear regression methods are usually used to screen key features that affect the catalytic function of biological enzymes. However, this method requires the analysis of hundreds of millions of biological enzyme features, which often takes a lot of time and human resources, and it is difficult to determine the key features that affect the catalytic function of biological enzymes in a short period of time. In addition, the linear regression method only considers the linear relationship between the independent variable and the dependent variable, ignoring the complex nonlinear relationship of the real biological enzyme environment. Summary of the invention

[0005] In order to overcome the above problems, the present invention provides a method, device, electronic device and storage medium for screening key features of bioenzyme catalytic function. The screening method provided by the present invention can screen out key features that affect the catalytic function of bioenzymes from hundreds of millions of bioenzyme structures, charges and electric field features, and use the Shapley additive interpretation algorithm to break the "black box" nature of the machine learning model, revealing the quantitative relationship between the key features of bioenzymes and catalytic functions, thereby providing guidance for the design of industrial bioenzymes with high catalytic activity and high selectivity.

[0006] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0007] The first aspect of the present invention provides a method for screening key characteristics of biological enzyme catalytic function, comprising:

[0008] (1) Obtaining enzyme characteristics and enzyme catalytic activity;

[0009] (2) Using the characteristics of biological enzymes and enzyme catalytic activity, a recursive feature elimination method based on the random forest algorithm was used to select key features of biological enzymes;

[0010] (3) Using the key features of the screened biological enzymes, the random forest model is trained to obtain a random forest model based on the optimal hyperparameter value;

[0011] (4) The key features of the sample biological enzyme to be tested are input into the random forest model based on the optimal hyperparameter value to obtain the predicted value of the catalytic activity of the sample biological enzyme, and the contribution value of each key feature to the catalytic activity is quantitatively calculated based on the predicted value of the catalytic activity.

[0012] In one or more embodiments, in step (1), the enzyme characteristics include the structure, charge and electric field characteristics of the enzyme conformation selected from the molecular dynamics simulation trajectory.

[0013] Preferably, the duration of the molecular dynamics simulation is 30 to 50 ns; and the number of conformations of each biological enzyme selected from the molecular dynamics simulation trajectory is greater than or equal to 20.

[0014] Preferably, the structural features of the enzyme conformation include bond length, bond angle and dihedral angle.

[0015] Further preferably, the bond length characteristic data, the bond angle characteristic data and the dihedral angle characteristic data are obtained by visualization software VMD.

[0016] Preferably, the charge signature data is acquired via Chemshell software.

[0017] Preferably, the electric field characteristic data is calculated according to Coulomb's law.

[0018] In one or more embodiments, in step (1), the enzyme catalytic activity is represented by the reaction energy barrier calculated by Chemshell.

[0019] In one or more embodiments, in step (2), the number of key features of the biological enzyme screened out is 2 to 5.

[0020] In one or more embodiments, in step (3), the process of training the random forest model includes: randomly selecting a part of the bioenzyme conformations containing the key features of the bioenzyme as a training set of the random forest model, and another part as a test set of the random forest model, and selecting a model parameter adjustment method of drawing a learning curve to adjust important hyperparameters of the random forest model, and gradually determining the random forest model with the optimal hyperparameter values ​​through the learning curve;

[0021] Among them, important hyperparameters include n_estimators, max_depth, min_samples_leaf, and min_samples_split.

[0022] Preferably, the coefficient of determination (R 2) to evaluate the performance of the random forest model.

[0023] Preferably, the coefficient of determination (R 2 ) is calculated as:

[0024]

[0025] Where N represents the total number of enzyme conformations, i represents the i-th enzyme conformation, and y i is the reaction energy barrier of the ith enzyme conformation, represents the reaction energy barrier predicted by the i-th enzyme conformation, y i Represents the average value of the energy barrier of the enzyme reaction.

[0026] In one or more embodiments, in step (4), the Shapley additive interpretation algorithm is used to quantitatively calculate the specific contribution of each key feature to the catalytic activity.

[0027] The second aspect of the present invention provides a device for screening key characteristics of biological enzyme catalytic function, comprising:

[0028] Bio-enzyme feature acquisition module, used to obtain bio-enzyme features and enzyme catalytic activity;

[0029] The feature selection module is used to select the key features of the enzymes by using the bioenzyme characteristics and enzyme catalytic activity and adopting the recursive feature elimination method based on the random forest algorithm;

[0030] The model training module is used to train the random forest model by screening out the key features of the biological enzyme and obtain the random forest model based on the optimal hyperparameter value;

[0031] The quantitative contribution value module is used to input the key features of the sample biological enzyme to be tested into the random forest model based on the optimal hyperparameter value, obtain the catalytic activity prediction value of the sample biological enzyme, and quantitatively calculate the contribution value of each key feature to the catalytic activity based on the catalytic activity prediction value.

[0032] A third aspect of the present invention provides an electronic device, comprising:

[0033] Memory, used to store computer programs;

[0034] A processor is used to execute the computer program to implement the steps of screening key characteristics of the catalytic function of the biological enzyme described in the first aspect.

[0035] A fourth aspect of the present invention provides a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of screening key characteristics of the catalytic function of the biological enzyme described in the first aspect are implemented.

[0036] The beneficial effects of the present invention are:

[0037] The present invention provides a method device, electronic device and storage medium for screening key features of bioenzyme catalytic function. The structure, charge and electric field characteristics of bioenzyme conformations in molecular dynamics simulation trajectories of different types of bioenzymes of the same type and their catalytic activity are obtained; the bioenzyme characteristics and enzyme catalytic activity are used to select features using a recursive feature elimination method based on a random forest algorithm to screen out key features of the bioenzyme; the screened bioenzyme conformation samples are used to train a random forest model to obtain a random forest model based on an optimal hyperparameter value; the trained random forest model not only inputs the key features of the sample bioenzyme to be tested into the random forest model based on the optimal hyperparameter value, but also obtains the catalytic activity prediction value of the sample bioenzyme, and quantitatively calculates the contribution value of each key feature to the catalytic activity based on the catalytic activity prediction value.

[0038] The present invention adopts a recursive feature elimination method based on a random forest algorithm to perform feature selection from hundreds of millions of biological enzyme features, which can not only consider the linear relationship between biological enzyme features and catalytic activity, but also capture the complex nonlinear relationship between the two.

[0039] The screening method provided by the present invention can screen out the key features that affect the catalytic function of biological enzymes from hundreds of millions of biological enzyme structures, charges and electric field characteristics, and use the Shapley additive interpretation algorithm to break the "black box" nature of the machine learning model, revealing the quantitative relationship between the key features of biological enzymes and the catalytic function, thereby providing guidance for the design of industrial biological enzymes with high catalytic activity and high selectivity. DETAILED DESCRIPTION

[0040] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0041] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0042] In order to enable those skilled in the art to more clearly understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below in conjunction with specific embodiments.

[0043] Example 1

[0044] A method for screening key characteristics of biological enzyme catalytic function, comprising:

[0045] (1) Obtaining enzyme characteristics and enzyme catalytic activity;

[0046] (2) Using the characteristics of biological enzymes and enzyme catalytic activity, a recursive feature elimination method based on the random forest algorithm was used to select key features of biological enzymes;

[0047] (3) Using the key features of the screened biological enzymes, the random forest model is trained to obtain a random forest model based on the optimal hyperparameter value;

[0048] (4) The key features of the sample biological enzyme to be tested are input into the random forest model based on the optimal hyperparameter value to obtain the predicted value of the catalytic activity of the sample biological enzyme, and the contribution value of each key feature to the catalytic activity is quantitatively calculated based on the predicted value of the catalytic activity.

[0049] Specifically, in step (1), the structure, charge, electric field characteristics and catalytic activity of the bioenzyme conformation in the molecular dynamics simulation trajectory of different types of bioenzymes of the same type are obtained. The duration of the bioenzyme molecular dynamics simulation is 30ns, and 20 enzyme conformations are randomly selected from the bioenzyme molecular dynamics simulation trajectory. Among them, the acquisition of structure, charge and electric field characteristics: the structural characteristics of the bioenzyme conformation include bond length, bond angle and dihedral angle, and the bond length characteristic data, bond angle characteristic data and dihedral angle characteristic data are obtained by the visualization software VMD; the charge characteristic data is obtained by the Chemshell software; the electric field characteristic data is calculated according to Coulomb's law. Acquisition of catalytic activity: the catalytic reaction energy barriers of 20 enzyme conformations are calculated using the Chemshell software, and the catalytic reaction activity of the bioenzyme is represented by the calculated rate-determining step reaction energy barrier.

[0050] In step (2), the recursive feature elimination method obtains the importance of each feature through the feature importances attribute in the random forest model, prunes the least important features from the existing feature set, and recursively repeats the process in the pruned set until a subset containing only 2 to 5 key features is obtained.

[0051] In step (3), the process of training the random forest model includes: randomly selecting 80% of the enzyme conformations that contain the key features of the enzyme as the training set of the random forest model, and the remaining 20% ​​as the test set of the random forest model, and using the model parameter adjustment method of drawing a learning curve to adjust the important hyperparameters of the random forest model, and gradually determining the random forest model with the optimal hyperparameter values ​​through the learning curve; wherein the important hyperparameters include n_estimators, max_depth, min_samples_leaf and min_samples_split.

[0052] After the random forest model parameters are adjusted, use the test set to test whether the trained model meets the requirements.

[0053] The determination coefficient (R 2 ) was used to evaluate the accuracy of the results of the random forest model, and the coefficient of determination (R 2 ) is calculated as:

[0054]

[0055] Where N represents the total number of enzyme conformations, i represents the i-th enzyme conformation, and y i is the reaction energy barrier of the ith enzyme conformation, represents the reaction energy barrier predicted by the i-th enzyme conformation, y i Represents the average value of the energy barrier of the enzyme reaction.

[0056] The accuracy of the results of the random forest model test set meets the requirements, and the process goes to step (4).

[0057] In step (4), the key features of the sample biological enzyme to be tested are input into the random forest model based on the optimal hyperparameter value to obtain the catalytic activity prediction value f(x i ), and the Shapley additive interpretation algorithm was used to quantitatively calculate the specific contribution of each key feature to the catalytic activity.

[0058] The larger the absolute value of the average Shapley value of each key feature, the greater the contribution value of the feature.

[0059] The formula for Shapley's additive interpretation algorithm is:

[0060]

[0061] Among them, x i represents the conformation of an enzyme with N key features, f(x i ) represents the predicted value of catalytic activity corresponding to each enzyme conformation, represents the expected value (base value) of the random forest model output on the data set, is feature j for enzyme conformation x i The Shapely value affected by the prediction result. The Shapely value is the contribution value.

[0062] Represents the Shapley value of each feature in each enzyme conformation, which is the weighted average of all possible combinations of variable subsets. The specific algorithm is as follows:

[0063]

[0064] In the formula, represents the Shapley value of feature j, S is a subset of features, x1, x2…x n represents each feature, |S| is the non-zero item in subset S, and f x (S) represents the predicted value of subset S.

[0065] It should be noted that the above-mentioned numerical values ​​can be determined according to specific prediction requirements, and can be adjusted according to needs in different embodiments, and are not limited to the above-mentioned exemplary numerical ranges.

[0066] Example 2

[0067] (1) Multiple enzyme conformations were randomly obtained from the molecular dynamics simulation trajectories of polyethylene terephthalate (PET) hydrolases IsPETase, FAST-PETase, BurPL, LCC, RgPETase, PET2 and HiC. The molecular dynamics simulation time of each enzyme was 30 ns, and 20 enzyme conformations were randomly selected from the molecular dynamics simulation trajectory of each enzyme, resulting in a total of 140 enzyme conformations. The catalytic reaction activity of each enzyme conformation was calculated using Chemshell software, and the catalytic reaction activity of the enzyme was represented by the calculated rate-determining step reaction energy barrier.

[0068] The structural, charge and electric field characteristic parameters of each enzyme conformation are obtained. The specific characteristic parameters include bond length data, bond angle data, dihedral angle data, charge data and electric field data. The bond length data, bond angle data and dihedral angle data are obtained through the visualization software VMD, the charge data is obtained through the Chemshell software, and the electric field data is calculated according to Coulomb's law.

[0069] (2) A recursive feature elimination method based on the random forest algorithm is used for feature selection. The random forest model is a traditional machine learning model without complex hyperparameter adjustment. The recursive feature elimination method obtains the importance of each feature through the feature importances attribute in the random forest model, prunes the least important features from the existing feature set, and recursively repeats the process in the pruned set until a subset containing only the key features bond length Ser@H1-His@N1 and charge His@N1 is obtained.

[0070] (3) Using the selected key features of the enzyme: bond length Ser@H1-His@N1 and charge His@N1, a random forest regression model was trained to obtain a random forest model based on the optimal hyperparameter value;

[0071] Specifically: 80% of the screened enzyme conformational features containing key features of the enzyme are randomly selected as the training set of the random forest model, and the remaining 20% ​​are used as the test set of the random forest model. The model parameter adjustment method of drawing a learning curve is used to adjust the important hyperparameters of the random forest model, and the random forest model with the optimal hyperparameter values ​​is gradually determined through the learning curve; among them, the important hyperparameters include n_estimators=43, max_depth=3, min_samples_leaf=3, min_samples_split=7.

[0072] After the random forest model parameters are adjusted, use the test set to test whether the trained model meets the requirements.

[0073] The determination coefficient (R 2 ) evaluates the accuracy of the results of the random forest model. In this embodiment, R 2 It is 0.63, indicating that the model after training on the test set meets the requirements.

[0074] (3) The key features of the sample enzyme to be tested were input into the random forest model based on the optimal hyperparameter value to obtain the predicted value of the catalytic activity of the sample enzyme, and the Shapley additive interpretation algorithm was used to quantitatively calculate the specific contribution of each key feature to the catalytic activity. The Shapley values ​​of the key feature bond length Ser@H1-His@N1 and charge His@N1 were 0.49 and -0.9, respectively.

[0075] Example 3

[0076] A device for screening key characteristics of biological enzyme catalytic function, comprising:

[0077] Bio-enzyme feature acquisition module, used to obtain bio-enzyme features and enzyme catalytic activity;

[0078] The feature selection module is used to select the key features of the enzymes by using the bioenzyme characteristics and enzyme catalytic activity and adopting the recursive feature elimination method based on the random forest algorithm;

[0079] The model training module is used to train the random forest model by screening out the key features of the biological enzyme and obtain the random forest model based on the optimal hyperparameter value;

[0080] The quantitative contribution value module is used to input the key features of the sample biological enzyme to be tested into the random forest model based on the optimal hyperparameter value, obtain the catalytic activity prediction value of the sample biological enzyme, and quantitatively calculate the contribution value of each key feature to the catalytic activity based on the catalytic activity prediction value.

[0081] Example 4

[0082] An electronic device, comprising:

[0083] Memory, used to store computer programs;

[0084] A processor is used to execute the computer program to implement the steps of screening key features of the bio-enzyme catalytic function in the above-mentioned embodiment 1; for example, including:

[0085] (1) Obtaining enzyme characteristics and enzyme catalytic activity;

[0086] (2) Using the characteristics of biological enzymes and enzyme catalytic activity, a recursive feature elimination method based on the random forest algorithm was used to select key features of biological enzymes;

[0087] (3) Using the key features of the screened biological enzymes, the random forest model is trained to obtain a random forest model based on the optimal hyperparameter value;

[0088] (4) The key features of the sample biological enzyme to be tested are input into the random forest model based on the optimal hyperparameter value to obtain the predicted value of the catalytic activity of the sample biological enzyme, and the contribution value of each key feature to the catalytic activity is quantitatively calculated based on the predicted value of the catalytic activity.

[0089] Example 5

[0090] A computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the steps of screening key features of the bioenzyme catalytic function in the above-mentioned embodiment 1; for example, including:

[0091] (1) Obtaining enzyme characteristics and enzyme catalytic activity;

[0092] (2) Using the characteristics of biological enzymes and enzyme catalytic activity, a recursive feature elimination method based on the random forest algorithm was used to select key features of biological enzymes;

[0093] (3) Using the key features of the screened biological enzymes, the random forest model is trained to obtain a random forest model based on the optimal hyperparameter value;

[0094] (4) The key features of the sample biological enzyme to be tested are input into the random forest model based on the optimal hyperparameter value to obtain the predicted value of the catalytic activity of the sample biological enzyme, and the contribution value of each key feature to the catalytic activity is quantitatively calculated based on the predicted value of the catalytic activity.

[0095] Although the above describes the specific implementation methods of the present invention, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A method for screening key characteristics of biological enzyme catalytic function, characterized in that: include: (1) Obtaining enzyme characteristics and enzyme catalytic activity; (2) Using the characteristics of biological enzymes and enzyme catalytic activity, a recursive feature elimination method based on the random forest algorithm was used to select key features of biological enzymes; (3) Using the key features of the screened biological enzymes, the random forest model is trained to obtain a random forest model based on the optimal hyperparameter value; (4) The key features of the sample biological enzyme to be tested are input into the random forest model based on the optimal hyperparameter value to obtain the predicted value of the catalytic activity of the sample biological enzyme, and the contribution value of each key feature to the catalytic activity is quantitatively calculated based on the predicted value of the catalytic activity.

2. The method for screening key characteristics of enzyme catalytic function according to claim 1, characterized in that: In step (1), the bioenzyme characteristics include the structure, charge and electric field characteristics of the bioenzyme conformation in the molecular dynamics simulation trajectory.

3. The method for screening key characteristics of biological enzyme catalytic function according to claim 2, characterized in that: The duration of the molecular dynamics simulation is 30 to 50 ns; the number of enzyme conformations selected from the molecular dynamics simulation trajectory is greater than or equal to 20; Or, the structural characteristics of the biological enzyme conformation include bond length, bond angle and dihedral angle; preferably, the bond length characteristic data, bond angle characteristic data and dihedral angle characteristic data are obtained by visualization software VMD; Alternatively, charge characteristic data were obtained using Chemshell software; Alternatively, the electric field characteristic data is calculated based on Coulomb's law.

4. The method for screening key characteristics of biological enzyme catalytic function according to claim 2, characterized in that: In step (1), the enzyme catalytic activity is represented by the reaction energy barrier calculated by Chemshell.

5. The method for screening key characteristics of biological enzyme catalytic function as described in claim 1, wherein in step (2), the number of key characteristics of the biological enzyme screened out is 2 to 5.

6. The method for screening key characteristics of biological enzyme catalytic function according to claim 5, characterized in that: In step (3), the process of training the random forest model includes: randomly selecting a part of the bioenzyme conformations containing the key features of the bioenzyme as a training set of the random forest model, and another part as a test set of the random forest model, using a model parameter adjustment method of drawing a learning curve to adjust important hyperparameters of the random forest model, and gradually determining the random forest model with the optimal hyperparameter values ​​through the learning curve; Among them, important hyperparameters include n_estimators, max_depth, min_samples_leaf, and min_samples_split.

7. The method for screening key characteristics of biological enzyme catalytic function according to claim 1, characterized in that: In step (4), the Shapley additive interpretation algorithm is used to quantitatively calculate the specific contribution of each key feature to the catalytic activity.

8. A device for screening key characteristics of biological enzyme catalytic function, characterized in that: include: Bio-enzyme feature acquisition module, used to obtain bio-enzyme features and enzyme catalytic activity; The feature selection module is used to select the key features of the enzymes by using the bioenzyme characteristics and enzyme catalytic activity and adopting the recursive feature elimination method based on the random forest algorithm; The model training module is used to train the random forest model by screening out the key features of the biological enzyme and obtain the random forest model based on the optimal hyperparameter value; The quantitative contribution value module is used to input the key features of the sample biological enzyme to be tested into the random forest model based on the optimal hyperparameter value, obtain the catalytic activity prediction value of the sample biological enzyme, and quantitatively calculate the contribution value of each key feature to the catalytic activity based on the catalytic activity prediction value.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the step of screening the key characteristics of the catalytic function of the biological enzyme as described in any one of claims 1 to 7.

10. A computer-readable storage medium for storing a computer program; characterized in that: When the computer program is executed by a processor, the step of screening the key characteristics of the catalytic function of the biological enzyme according to any one of claims 1 to 7 is implemented.

Images