Method for predicting interaction relationship among aroma substances based on decision tree algorithm

Abstract

The invention discloses a method for predicting an interaction relationship between aroma substances based on a decision tree algorithm. The method comprises the following steps: constructing an aroma substance interaction relationship database; training the C5.0 decision tree prediction model based on the Booting algorithm by taking data of the aroma substance interaction relationship database as a training test data set to obtain an optimal prediction model; and utilizing the optimal prediction model to obtain the interaction relationship of the aroma substance combination to be predicted. According to the method disclosed by the invention, the decision tree model is adopted to carry out association modeling on binding energy data and sensory experiment results, and a Boosting algorithm is introduced to optimize model parameters, so that deep fusion of olfactory receptor binding energy and sensory evaluation data is realized; and through multi-dimensional data integration, the aroma substance interaction prediction precision is remarkably improved, and the method has wide applicability and good application prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of food flavor chemistry, and relates to a method for predicting the interaction relationship between two different aroma compounds in food, and particularly relates to a method for predicting the interaction relationship between aroma substances based on a decision tree algorithm. Background Art

[0002] At present, the research methods for the interaction between aroma substances mainly include the threshold method, the Feller additive model method, the OAV method, and the σ-τ diagram method. However, the above methods are all at the level of sensory evaluation and are affected by a large number of subjective factors (such as environmental temperature and humidity), with many uncertain factors and a lack of objective analysis and verification. There is also a method that simply judges the interaction relationship by directly observing the binding energy generated after the interaction between one substance and two substances with olfactory receptors. Although it can study the interaction between aroma substances from the data level, its operation is relatively complex and lacks detailed data analysis, making it difficult to ensure the reliability of the evaluation.

[0003] Therefore, it is of great practical significance to develop a method for predicting the interaction relationship between aroma substances that is relatively objective and has high prediction accuracy. Summary of the Invention

[0004] Due to the above-mentioned defects in the prior art, the present invention provides a method for predicting the interaction relationship between aroma substances that is relatively objective and has high prediction accuracy. Specifically, it is a method for predicting the interaction relationship between aroma substances based on a decision tree algorithm, which overcomes the defect that a single sensory evaluation method lacks objective analysis, and the defect that the binding energy judgment is too simple and lacks a calculation basis.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A method for predicting the interaction relationship between aroma substances based on a decision tree algorithm, comprising the following steps:

[0007] (1) Select the types of aroma substances, separately dock different monomer aroma substances with olfactory receptors OR1A1 and OR2W1 to obtain initial binding energy data. Introduce exogenous monomer aroma substances on the basis of the monomer aroma substance-olfactory receptor complex for secondary docking to obtain composite binding energy data. At the same time, determine the interaction relationship between the selected aroma substances in pairs through artificial sensory experiments, and use the initial binding energy data, composite binding energy data, and the interaction relationship between the aroma substances obtained from the corresponding artificial sensory experiments as a set of data to construct a database of the interaction relationship between aroma substances;

[0008] (2)Using the data in the aroma substance interaction relationship database constructed in step (1) as the training and test data set to train the C5.0 decision tree prediction model based on the Booting algorithm, an optimal prediction model is obtained. The input of the C5.0 decision tree prediction model based on the Booting algorithm is the initial binding energy data and the composite binding energy data, and the output is the interaction relationship between aroma substances obtained from the artificial sensory experiment;

[0009] (3)Separate the monomer aroma substances in the aroma substance combination to be predicted from the olfactory receptors OR1A1 and OR2W1 for individual docking to obtain the initial test binding energy data. On the basis of the monomer aroma substance-olfactory receptor complex, introduce another monomer aroma substance in the aroma substance combination to be predicted for secondary docking to obtain the composite test binding energy data. The monomer aroma substances in the aroma substance combination to be predicted are all included in the aroma substances selected in step (1);

[0010] (4)Input the initial test binding energy data and the composite test binding energy data obtained in step (3) into the optimal prediction model, and the optimal prediction model outputs the interaction relationship of the aroma substance combination to be predicted.

[0011] C5.0 is an optimized algorithm based on the classical decision tree classification algorithm C4.5 and belongs to the category of machine learning supervised learning. It can be used to solve classification problems. It was extended and optimized from the ID3 algorithm by Ross Quinlan in 1993. It selects features through the gain ratio, uses the dichotomy and multi-method to process continuous features, and uses the majority voting method or the probability distribution method to process missing values, enhancing the robustness of the algorithm.

[0012] To further improve the prediction accuracy of the decision tree model, the present invention adopts the C5.0 decision tree prediction model based on the Booting algorithm, that is, uses the Booting algorithm to improve the C5.0 decision tree prediction algorithm, which can further improve the prediction accuracy. At the same time, the binding energy observation of a single aroma substance-receptor is extended to the energy analysis of the composite aroma system, and the cross-validation of multi-dimensional data significantly improves the accuracy of the determination of the aroma interaction type. At the same time, the model has good extrapolation and can be applied to predict the interaction mode of other aroma combinations in the same matrix environment (the same solution system).

[0013] The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm of the present invention performs a binding energy test on the aroma substance combination to be predicted. After obtaining relevant data, it inputs the data into the C5.0 decision tree prediction model based on the Booting algorithm to obtain the interaction relationship between aroma substances for the aroma substance combination to be predicted. By screening out important factors affecting the interaction relationship between aromas through the decision tree model and finally predicting the interaction result, the overall accuracy rate reaches 95.31%. The present invention provides ideas and methods for quickly and accurately predicting the interaction effects between aroma substances using machine learning, and has good application prospects.

[0014] As a preferred technical solution:

[0015] For the method for predicting the interaction relationship between aroma substances based on the decision tree algorithm as described above, the interaction relationships between the selected aroma substances in pairs are obtained by using the same sensory experiment method.

[0016] For the method for predicting the interaction relationship between aroma substances based on the decision tree algorithm as described above, the same sensory experiment method is the Feller additive model method.

[0017] For the method for predicting the interaction relationship between aroma substances based on the decision tree algorithm as described above, all artificial sensory experiments are based on the same solution mechanism. Experimental results obtained under different mechanisms will be different, which will affect the accuracy of the collected database.

[0018] For the method for predicting the interaction relationship between aroma substances based on the decision tree algorithm as described above, the single docking and secondary docking are completed using the AutoDock Vina software.

[0019] For the method for predicting the interaction relationship between aroma substances based on the decision tree algorithm as described above, the C5.0 decision tree prediction model based on the Booting algorithm is the C5.0 decision tree prediction model of AdaBoost integrated with 10 trees. Boosting ensemble learning corrects the errors of the previous model step by step through multiple rounds of iteration, constructs a strong learner, significantly improves the performance of the original decision tree model, and significantly improves the overall accuracy probability of predicting the interaction effects between substances.

[0020] For the method for predicting the interaction relationship between aroma substances based on the decision tree algorithm as described above, the input of the C5.0 decision tree prediction model based on the Booting algorithm is a continuous measurement value composed of initial binding energy data and composite binding energy data.

[0021] The above technical solutions are only one feasible technical solution of the present invention, and the protection scope of the present invention is not limited thereto. Those skilled in the art can reasonably adjust the specific design according to actual needs.

[0022] The above invention has the following advantages or beneficial effects:

[0023] (1) The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm of the present invention constructs an analysis system including three interaction types of synergism, addition, and masking based on artificial sensory experiments and molecular docking technology (judging the interaction effect by the binding energy of aroma substance-olfactory receptor).

[0024] (2) The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm of the present invention uses a decision tree model to perform correlation modeling on the binding energy data and the results of sensory experiments, and introduces the Boosting algorithm to optimize the model parameters, realizing the deep integration of olfactory receptor binding energy and sensory evaluation data.

[0025] (3) The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm of the present invention significantly improves the prediction accuracy of the interaction between aroma substances through multi-dimensional data integration, and has the technical advantages of simple operation and high operation efficiency. Its visual output results are intuitive and reliable and have wide applicability, with good application prospects. Brief Description of the Drawings

[0026] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, the present invention and its features, shape, and advantages will become more obvious. The same reference numerals indicate the same parts in all the drawings. The drawings are not drawn to scale, and the focus is on showing the gist of the present invention.

[0027] Figure 1 It is a step sequence diagram of the method for predicting the interaction relationship between aroma substances based on the decision tree algorithm of the present invention. Detailed Embodiments

[0028] The present invention will be further described below in conjunction with the drawings and specific embodiments, but it is not limited to the present invention.

[0029] The molecular docking method, sensory experiment results, and monomer aroma compound selection involved in the following examples and comparative examples are based on the following literature records, specifically as follows: Research on the Interaction of Different Aroma Substances Based on S Curve and Molecular Docking Technology [J]. China Food Additives, 2023, 34(06): 119-129.

[0030] Example 1

[0031] A method for predicting the interaction relationship between aroma substances based on the decision tree algorithm, the steps are as follows (as Figure 1 shown):

[0032] (1) Construct a database of the interaction relationship between aroma substances, specifically:

[0033] Nine aroma substances were selected, and molecular docking experiments were performed using AutoDock Vina software. Independent docking of monomeric aroma compounds with olfactory receptors OR1A1 and OR2W1 was carried out to obtain initial binding energy data. Subsequently, exogenous monomers were introduced on the basis of the monomer-receptor complex for secondary docking to obtain complex binding energy data. At the same time, the interaction relationships (synergy / addition / masking) between the selected aroma substances were determined through artificial sensory experiments. To ensure the reliability of the database, all sensory experiments adopted a standardized process: unified sensory evaluation method (Feller additive model method) and solution system. Using the initial binding energy data, complex binding energy data, and the interaction relationships between the aroma substances obtained from the corresponding artificial sensory experiments as a set of data, an aroma substance interaction relationship database was constructed (using the initial binding energy data and complex binding energy data as input variables of the database, and the aroma interaction relationship (synergy / addition / masking) determined by artificial sensory experiments as the target variable). Finally, a prediction database containing 64 sets of valid data was constructed. The specific types of aroma substances, input variables, and target variable data are shown in Table 1 below. In Table 1, 1 represents a synergistic effect, 2 represents an additive effect, and 3 represents a masking effect;

[0034] Table 1

[0035]

[0036] (2) Using the data of the aroma substance interaction relationship database constructed in step (1) as the training and test data set, the C5.0 decision tree prediction model based on the Booting algorithm was trained to obtain the optimal prediction model. The input (i.e., input variables) of the C5.0 decision tree prediction model based on the Booting algorithm was the initial binding energy data and complex binding energy data (specifically, the docking binding energy of aroma substance A with receptor protein OR1A1 / OR2W1, the docking binding energy of aroma substance B with receptor protein OR1A1 / OR2W1, the docking binding energy of aroma substance A + receptor protein OR1A1 / OR2W1 with exogenous aroma substance B, and the docking binding energy of aroma substance B + receptor protein OR1A1 / OR2W1 with exogenous aroma substance A), and the output (i.e., target variable) was the interaction relationship between the aroma substances obtained from artificial sensory experiments. Six input variables were set as continuous data, and one target variable was set as nominal data for classification;

[0037] (3) Individually dock the monomer aroma substances in the aroma substance combination to be predicted (i.e., the aroma substance combination in the aroma substance interaction relationship database) with olfactory receptors OR1A1 and OR2W1 to obtain initial test binding energy data. Based on the monomer aroma substance - olfactory receptor complex, introduce another monomer aroma substance in the aroma substance combination to be predicted for secondary docking to obtain composite test binding energy data. The monomer aroma substances in the aroma substance combination to be predicted are all included in the aroma substances selected in step (1);

[0038] (4) Input the initial test binding energy data and composite test binding energy data obtained in step (3) into the optimal prediction model. The optimal prediction model outputs the interaction relationship of the aroma substance combination to be predicted. The model output results are shown in Table 2;

[0039] Table 2

[0040]

[0041] As can be seen from the results in Table 2, the overall prediction accuracy of the optimal prediction model reaches 95.31%. Among the 64 groups of data, 61 groups of data are predicted correctly, and only 3 groups of data are predicted wrongly.

[0042] Comparative Example 1

[0043] A method for predicting the interaction relationship between aroma substances, which is basically the same as Example 1, except that the C5.0 decision tree prediction model based on the Booting algorithm in Example 1 is replaced with a C5.0 decision tree prediction model. The model output results are shown in Table 3. Among the 64 groups of data, 54 groups of data results are predicted correctly, and 10 groups of data results are predicted wrongly. The overall prediction accuracy rate is only 84.38%.

[0044] Table 3

[0045]

[0046] By comparing Example 1 and Comparative Example 1, it can be found that integrating the Boosting technology on the basis of the original decision tree improves the attention of the overall model to a few unbalanced samples, optimizes the problem of data processing imbalance, serially trains the original decision tree model, corrects the errors of the previous model, and improves the overall prediction accuracy rate.

[0047] Comparative Example 2

[0048] A method for predicting the interaction relationship between aroma substances is basically the same as that in Example 1, except that the C5.0 decision tree prediction model based on the Booting algorithm in step (2) takes the docking binding energy of aroma substance A and receptor protein OR1A1 / OR2W1 and the docking binding energy of aroma substance B and receptor protein OR1A1 / OR2W1 as inputs, and the model output results are shown in Table 4. Among the 64 groups of data, 59 groups of data results are predicted correctly and 5 groups of data results are predicted wrongly, and the overall accuracy rate is 92.19%.

[0049] Table 4

[0050]

[0051] By comparing Example 1 and Comparative Example 2, it can be found that by introducing the dual-source data comparison mechanism of the composite system, the observation of the binding energy of a single aroma substance-receptor is extended to the energy analysis of the composite aroma system, and the accuracy of determining the type of aroma interaction is significantly improved by combining multi-dimensional data cross-validation.

[0052] In summary, the method for predicting the interaction relationship between aroma substances based on the decision tree model of the present invention is simple and fast to operate, and the results are intuitive and reliable.

[0053] Those skilled in the art should understand that those skilled in the art can implement variations in combination with the prior art and the above embodiments, which will not be elaborated here. Such variations do not affect the essence of the present invention and will not be elaborated here.

[0054] The preferred embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and the equipment and structures not described in detail should be understood to be implemented in a common manner in the art; any person skilled in the art can make many possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the scope of the technical solution of the present invention, or modify it into an equivalent embodiment with equivalent changes, which does not affect the essence of the present invention. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the protection of the technical solution of the present invention.

Claims

1. A method for predicting the interaction relationship between aroma substances based on a decision tree algorithm, characterized in that: It includes the following steps: (1) Select the types of aroma substances. Separate individual aroma substances are separately docked with olfactory receptors OR1A1 and OR2W1 to obtain initial binding energy data. Based on the monomer aroma substance-olfactory receptor complex, an exogenous monomer aroma substance is introduced for secondary docking to obtain composite binding energy data. At the same time, the interaction relationships between the selected aroma substances are determined through artificial sensory experiments. Using the initial binding energy data, composite binding energy data, and the interaction relationships between the aroma substances obtained from the corresponding artificial sensory experiments as a set of data, a database of aroma substance interaction relationships is constructed; (2) Using the data in the database of aroma substance interaction relationships constructed in step (1) as the training and test data set to train the C5.0 decision tree prediction model based on the Booting algorithm to obtain the optimal prediction model. The input of the C5.0 decision tree prediction model based on the Booting algorithm is the initial binding energy data and the composite binding energy data, and the output is the interaction relationships between the aroma substances obtained from artificial sensory experiments; (3) Separate individual aroma substances in the aroma substance combination to be predicted are separately docked with olfactory receptors OR1A1 and OR2W1 to obtain initial test binding energy data. Based on the monomer aroma substance-olfactory receptor complex, another monomer aroma substance in the aroma substance combination to be predicted is introduced for secondary docking to obtain composite test binding energy data. The monomer aroma substances in the aroma substance combination to be predicted are all included in the aroma substances selected in step (1); (4) Input the initial test binding energy data and the composite test binding energy data obtained in step (3) into the optimal prediction model, and the optimal prediction model outputs the interaction relationships of the aroma substance combination to be predicted.

2. The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm according to claim 1, wherein The interaction relationships between the selected aroma substances are obtained by using the same sensory experiment method.

3. The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm according to claim 2, wherein The same sensory experiment method is the Feller additive model method.

4. The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm according to claim 2, characterized in that, All artificial sensory experiments are of the same solution mechanism.

5. The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm according to claim 1, wherein The separate docking and secondary docking are completed using the AutoDock Vina software.

6. The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm according to claim 1, characterized in that The C5.0 decision tree prediction model based on the Booting algorithm is a C5.0 decision tree prediction model integrated with AdaBoost.

7. The method for predicting the interaction relationship between aroma substances based on the decision tree algorithm according to claim 1, characterized in that The input of the C5.0 decision tree prediction model based on the Booting algorithm is a continuous measurement value composed of the initial binding energy data and the composite binding energy data.

Images