Prediction method, system, equipment and medium for inhibition specificity of antibacterial peptide to gram-positive bacteria and gram-negative bacteria

Abstract

The invention belongs to the technical field of biological high-throughput data analysis, and particularly relates to a method, a system, equipment and a medium for predicting inhibition specificity of an antibacterial peptide to gram-positive bacteria and gram-negative bacteria, the method comprises the following steps: acquiring an antibacterial peptide sequence with known activity, and respectively sorting antibacterial peptide data sets with gram-positive bacteria activity and gram-negative bacteria activity; on the basis of the obtained antibacterial peptide sequence, three feature models are constructed, the three feature models are trained and predicted through multiple machine learning algorithms, and the optimal model is screened through cross validation. According to the method, by introducing a multi-source feature construction and multi-model integration evaluation mechanism, the limitation that a traditional method depends on a single feature or a single algorithm is broken through, and accurate prediction of the activity difference of the antibacterial peptides between gram-positive bacteria and gram-negative bacteria can be achieved. Besides, the method not only can be used for classifying and predicting the activity of known antibacterial peptides, but also can be used for excavating potential novel antibacterial peptides which are not recognized in the past and have specific activity, and has a wide application prospect.

Description

Technical Field

[0001] The present invention belongs to the technical field of biological high-throughput data analysis, and in particular relates to a method, system, equipment and medium for predicting the inhibitory specificity of antimicrobial peptides to Gram-positive and Gram-negative bacteria. Background Art

[0002] Currently, studies have used machine learning methods to predict the inhibitory activity of antimicrobial peptides (AMPs) against Gram-positive bacteria (such as Staphylococcus aureus) and Gram-negative bacteria (such as Escherichia coli). For example, Söykenez et al. (2022) proposed a prediction model based on algorithms such as AdaBoost, decision trees, and SVMs, and evaluated performance through Monte Carlo cross-validation (MCCV). However, the code was not publicly available and lacked experimental validation, making the results difficult to reproduce. Furthermore, the DBAASP database provides tools for predicting AMP activity, but only supports the upload and analysis of a small number of sequences (e.g., 25) at a time, which is insufficient for high-throughput screening. Furthermore, the prediction results have not been experimentally validated, raising questions about their reliability. Existing technologies often rely on single features (such as sequence composition or physicochemical properties) or single algorithms (such as SVMs), resulting in limited prediction accuracy and generalization capabilities.

[0003] More specifically, there is the problem of overly simplistic feature extraction. Most studies only use amino acid composition or simple physical and chemical parameters, which fail to comprehensively reflect the complex mechanisms of interaction between antimicrobial peptides and different bacterial species. Secondly, the model construction lacks systematic optimization. Researchers often use a single machine learning algorithm and do not fully evaluate the adaptability of different feature combinations and algorithm architectures, resulting in difficulty in achieving breakthroughs in prediction performance. In addition, existing technology platforms generally have insufficient processing power to meet the needs of large-scale antimicrobial peptide screening, which seriously restricts their application value in drug development. These defects together lead to the fact that the accuracy and practicality of existing antimicrobial peptide activity prediction methods are difficult to meet the actual scientific research and industrial needs. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, system, equipment and medium for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria, so as to improve the accuracy and practicality of predicting the inhibitory activity of antimicrobial peptides against Gram-positive and Gram-negative bacteria, and to overcome the defects of existing methods such as single features, insufficient model optimization and lack of experimental verification.

[0005] The present invention achieves the above-mentioned purpose through the following technical solutions: In a first aspect, the present invention provides a method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria, the method comprising: Obtaining antimicrobial peptide sequences to be processed that have antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria respectively; The antimicrobial peptide sequence to be processed is segmented to obtain a multi-dimensional frequency feature matrix; The physical and chemical characteristics of the antimicrobial peptide sequence to be processed are extracted to obtain a physical and chemical characteristic matrix; splicing the frequency characteristic matrix and the physical and chemical characteristic matrix to form a composite characteristic matrix; Multiple feature models built based on machine learning algorithms are trained using frequency feature matrices, physical and chemical feature matrices, and composite feature matrices as model inputs; Determine the target prediction model based on the excellence evaluation results of multiple feature models after training; The antimicrobial peptide sequences to be tested are predicted based on the target prediction model to obtain the specific classification results of their inhibition on Gram-positive bacteria and / or Gram-negative bacteria.

[0006] Furthermore, the method of obtaining the antimicrobial peptide sequence to be processed that has antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria comprises: Obtaining antimicrobial peptide sequences that have antimicrobial activity against Gram-positive and Gram-negative bacteria; Adding a marker with inhibitory activity at a set position in the sequence; The repeated sequences in the antimicrobial peptide sequence are removed to obtain the antimicrobial peptide sequence to be processed.

[0007] Furthermore, the antimicrobial peptide sequence to be processed is segmented to obtain a multi-dimensional frequency feature matrix, including: Sliding cuts are performed on the short peptide chain composed of amino acids in the antimicrobial peptide sequence to be processed, with three consecutive amino acids as a unit, to generate multiple tripeptide combinations; The occurrence frequencies of all tripeptide combinations in the antimicrobial peptide sequence to be processed are counted to construct a multidimensional frequency feature matrix, wherein each dimension in the frequency feature matrix corresponds to the frequency value of a tripeptide combination.

[0008] Furthermore, the physicochemical characteristics of the antimicrobial peptide sequences to be processed were extracted, including the following extraction indicators: isoelectric point, sequence length, molecular weight, hydrophobicity, hydrophobic moment, disordered conformation tilt, hydrophilicity index and net charge.

[0009] Furthermore, the multiple feature models constructed based on the machine learning algorithm are trained using the frequency feature matrix, the physical and chemical feature matrix, and the composite feature matrix as model inputs, respectively, including: Construct frequency feature matrix models, physical and chemical property feature models, and feature splicing models based on machine learning algorithms; The frequency feature matrix, the physical and chemical feature matrix, and the composite feature matrix are respectively input into corresponding models as training sets for training; The machine learning algorithms used include decision trees, random forests, support vector machines, conditional inference trees, and naive Bayes.

[0010] Furthermore, determining the target prediction model according to the excellence evaluation results of the multiple feature models after training includes: Use cross-validation to evaluate the set evaluation indicators of each machine learning algorithm under different feature models; Based on the evaluation results, the best performing model and feature combination is selected as the target prediction model; The set evaluation indicators include precision, recall and Kappa coefficient.

[0011] Furthermore, the model output is the specific inhibition result of the antimicrobial peptide against Gram-positive bacteria and / or Gram-negative bacteria, including sequence ID and predicted specific classification result.

[0012] In a second aspect, the present invention provides a prediction system for the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria, which is used to perform the above-mentioned prediction method, and the system comprises: A data acquisition module is used to obtain antimicrobial peptide sequences to be processed that have antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria respectively; The fragment segmentation module is used to segment the antimicrobial peptide sequence to be processed and obtain a multi-dimensional frequency feature matrix; A feature extraction module is used to extract the physical and chemical features of the antimicrobial peptide sequence to be processed and obtain a physical and chemical feature matrix; A feature splicing module, used for splicing the frequency feature matrix and the physical and chemical feature matrix to form a composite feature matrix; The model training module is used to train multiple feature models built based on machine learning algorithms, using frequency feature matrices, physical and chemical feature matrices, and composite feature matrices as model inputs. A model determination module is used to determine the target prediction model based on the excellence evaluation results of multiple feature models after training; The prediction module is used to predict the antimicrobial peptide sequence to be tested based on the target prediction model to obtain its inhibition specificity classification results on Gram-positive bacteria and / or Gram-negative bacteria.

[0013] In a third aspect, the present invention provides a computer device, comprising: a processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the prediction method as described above.

[0014] In a fourth aspect, the present invention proposes a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a processor of a computer device, the computer device is enabled to perform the prediction method as described above.

[0015] The beneficial effects of the present invention are: 1. This invention significantly improves the accuracy and practicality of predicting the inhibitory activity of antimicrobial peptides through multidimensional feature fusion and machine learning algorithm optimization. By integrating 3-mer frequency features with eight key physicochemical properties to construct a composite feature matrix, it comprehensively characterizes the structural and functional properties of antimicrobial peptides, overcoming the problem of single-feature methods in traditional methods.

[0016] 2. The present invention uses five algorithms, including random forest and support vector machines, for parallel training. Through 10-fold cross-validation, the model performance under different feature combinations is compared. Finally, the random forest model based on the physical and chemical feature matrix is ​​determined to be the optimal solution. The prediction accuracy reaches 77.78% (positive bacteria) and 55.56% (negative bacteria), and the overall accuracy is 66.67%, which is a significant improvement over the existing methods.

[0017] 3. This invention establishes a closed-loop "prediction-synthesis-verification" research system. The prediction results are validated through in vitro antibacterial experiments against Escherichia coli and Staphylococcus aureus, ensuring the reliability of the technical solution. Furthermore, this method supports high-throughput data analysis and can simultaneously process large numbers of sequences, providing an efficient and reliable technical platform for the rational design and drug development of antimicrobial peptides. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a schematic flow chart of a method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to one embodiment of the present application; Figure 2 This is another flow chart of a method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to one embodiment of the present application; Figure 3 This is a schematic diagram of various algorithm indicators of a model according to an experimental case in an implementation scenario of the specific embodiment of this application; Figure 4 This is a schematic diagram of the algorithm indicators of Model 2 according to an experimental case in an implementation scenario of the specific embodiment of this application; Figure 5 This is a schematic diagram of the algorithm indicators of Model 3 according to an experimental case in an implementation scenario of the specific embodiment of this application; Figure 6 This is a schematic diagram of the structure of a prediction system for the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to one embodiment of the present application; Figure 7 This is a heat map showing the difference in inhibition rates of various antimicrobial peptides against Escherichia coli and Staphylococcus aureus according to an experimental case in an implementation scenario in a specific embodiment of the present application; Figure 8This is a schematic structural diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0019] The present application is described in further detail below in conjunction with the accompanying drawings. It is necessary to point out that the following specific implementation methods are only used to further illustrate the present application and cannot be understood as limiting the scope of protection of the present application. Technicians in this field can make some non-essential improvements and adjustments to the present application based on the above application content.

[0020] In the prior art, the physicochemical properties of Gram-positive bacteria, Gram-negative bacteria, and antimicrobial peptides (AMPs) are described as follows: Gram-positive and Gram-negative bacteria differ significantly in their cell wall structure: Gram-positive bacteria have a thicker peptidoglycan layer and lack an outer membrane; Gram-negative bacteria, on the other hand, have a thinner peptidoglycan inner layer surrounded by a complex outer membrane containing lipopolysaccharides, phospholipids, membrane proteins, porins, and other components, which, to a certain extent, constitutes a natural barrier to drug resistance. These differences in cell wall structure significantly influence the mechanism of action and efficacy of drugs such as antibiotics and antimicrobial peptides.

[0021] Antimicrobial peptides (AMPs) have become a hot topic in the development of new antimicrobial drugs due to their antimicrobial activity and low drug resistance. AMPs typically act by disrupting bacterial membrane structure, forming pores, or disrupting membrane potential. Because Gram-positive bacteria lack an outer membrane, AMPs can more easily penetrate their thick peptidoglycan layer and interact directly with the cell membrane, resulting in a more pronounced antimicrobial effect against Gram-positive bacteria. However, the outer membrane of Gram-negative bacteria can partially prevent AMPs from approaching and penetrating, resulting in some antimicrobial peptides exhibiting lower activity against Gram-negative bacteria.

[0022] like Figure 1-2 As shown, in one embodiment, a method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive bacteria and Gram-negative bacteria is provided, and the method comprises the following steps: S1. Obtain the antimicrobial peptide sequence to be processed that has antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria.

[0023] Furthermore, obtaining an antimicrobial peptide sequence to be processed that has antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria, respectively, includes: obtaining an antimicrobial peptide sequence that has antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria, respectively; adding a marker with inhibitory activity, such as a sequence ID, at a set position in the sequence, so that the activity information of the antimicrobial peptide can be accurately identified during subsequent analysis and model training; and removing repeated sequences in the antimicrobial peptide sequence to obtain the antimicrobial peptide sequence to be processed.

[0024] In this embodiment, the antimicrobial peptide sequences to be processed that have antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria can be downloaded from the public antimicrobial peptide database DRAMP (Data repository of antimicrobial peptides). A program is written to perform redundancy processing on the raw data, remove duplicate sequences, and extract antimicrobial peptide sequences that have antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria separately.

[0025] It should be noted that in this example, active antimicrobial peptide sequences were downloaded from a public database to extract antimicrobial peptide sequences that have antimicrobial activity against Gram-positive bacteria or Gram-negative bacteria alone.

[0026] S2. Segment the antimicrobial peptide sequence to be processed to obtain a multi-dimensional frequency feature matrix.

[0027] Furthermore, the antimicrobial peptide sequence to be processed is segmented to obtain a multidimensional frequency feature matrix, which in an optional embodiment is specifically a 3-mer (tripeptide) frequency feature matrix, including: performing sliding cutting on a short peptide chain composed of amino acids in the antimicrobial peptide sequence to be processed with three consecutive amino acids as a unit to generate multiple tripeptide combinations and construct a 3-mer frequency feature matrix; and counting the occurrence frequencies of all tripeptide combinations in the antimicrobial peptide sequence to be processed to construct a multidimensional frequency feature matrix, wherein each dimension in the frequency feature matrix corresponds to the frequency value of a tripeptide combination.

[0028] Exemplarily, a short peptide chain composed of amino acids in the antimicrobial peptide sequence to be processed is subjected to sliding cleavage, with three consecutive amino acids as a unit, to generate multiple tripeptide combinations. For example, for the sequence "ABC," sliding cleavage can produce the tripeptide combination "ABC"; for the sequence "ABCD," sliding cleavage can produce two tripeptide combinations, "ABC" and "BCD." For example, if any three of the 20 amino acids are randomly selected and arranged (repeats are allowed), there are 8,000 unique tripeptide combinations, forming an 8,000-dimensional vector in the frequency feature matrix. The frequency of occurrence of all tripeptide combinations in the antimicrobial peptide sequence is counted to construct a multidimensional frequency feature matrix, in which each dimension corresponds to the frequency value of a tripeptide combination, forming a multidimensional frequency feature matrix.

[0029] S3. Extract the physicochemical characteristics of the antimicrobial peptide sequence to be processed to obtain a physicochemical characteristic matrix.

[0030] Furthermore, the physicochemical characteristics of the antimicrobial peptide sequences to be processed were extracted, including the following extraction indicators: isoelectric point, sequence length, molecular weight, hydrophobicity, hydrophobic moment, disordered conformation tilt, hydrophilicity index and net charge.

[0031] S4. Concatenate the frequency feature matrix and the physical and chemical feature matrix to form a composite feature matrix.

[0032] More specifically, the above two types of feature matrices are spliced ​​column by column to construct a composite feature matrix, in which each antimicrobial peptide sequence corresponds to a complete feature vector for subsequent training and prediction.

[0033] For example, the composite feature matrix formed by splicing in step S4 is as follows: For the 3mer amino acid matrix, the dictionary function of Python in the prior art is used, where the key value is the sequence, and the value value is the number of amino acid fragments corresponding to the sequence in the multidimensional frequency feature matrix. For example, for the sequence "ABC", its key is "ABC", and the value value is all the numerical values ​​of the sequence "ABC" in this row in the multidimensional frequency feature matrix. For the physicochemical feature matrix, the key value is the same sequence, but the value value is the physicochemical feature value corresponding to the sequence in the physicochemical feature matrix. For example, for the sequence "ABC", its key is "ABC", and the value value is all the numerical values ​​of the sequence "ABC" in this row in the physicochemical feature matrix. Then, the multidimensional frequency and physicochemical feature values ​​with the same key value, that is, the same sequence, can be combined to form a composite feature matrix.

[0034] As shown in Table 1, the multidimensional frequency feature matrix, physicochemical feature matrix, and composite feature matrix in this embodiment can all be represented by Table 1, but the features represented by feature 1, feature 2, feature 3, and feature 4 in different matrices are different. In the multidimensional frequency feature matrix, feature 1, feature 2, feature 3, and feature 4 are all tripeptide combinations. In the physicochemical feature matrix, feature 1, feature 2, feature 3, and feature 4 are all physicochemical features. In the composite feature matrix, feature 1, feature 2, and feature 3 are tripeptide combinations, and feature 4 is a physicochemical feature.

[0035] Table 1 Eigenvalue matrix Sequence ID Feature 1 Feature 2 Feature 3 Feature 4 Sequence1 17 261 11 4 Sequence 2 28 80 13 12 Sequence 3 14 21 6 15 Sequence 4 1 42 2 1 Sequence5 2 8 3 5 Sequence 6 0 7 4 0 Sequence 7 1 1 0 2 In a specific embodiment, the method further includes dividing the multidimensional frequency feature matrix, physical and chemical feature matrix, and composite feature matrix obtained above into a training group and a test group, as shown in Table 2.

[0036] Table 2 Group list Sequence ID Antibacterial activity Group Sequence1 Positive Training group / test group Sequence 2 Negative Training group / test group Sequence 3 Negative Training group / test group Sequence 4 Positive Training group / test group S5. Multiple feature models built based on machine learning algorithms are trained using frequency feature matrix, physical and chemical feature matrix, and composite feature matrix as model inputs.

[0037] Furthermore, multiple feature models constructed based on machine learning algorithms are trained using frequency feature matrices, physical and chemical feature matrices, and composite feature matrices as model inputs, including: constructing frequency feature matrix models, physical and chemical property feature models, and feature splicing models based on machine learning algorithms; inputting frequency feature matrices, physical and chemical feature matrices, and composite feature matrices as training sets into corresponding models for training; and machine learning algorithms using decision trees, random forests, support vector machines (SVMs), conditional inference trees (Ctrees), and naive Bayes. In this application, a random forest algorithm was used to train a model predicting the inhibitory specificity of antimicrobial peptides. This algorithm completes the classification task by constructing multiple decision trees and performing ensemble voting. The specific implementation process involves using bootstrap sampling with replacement to extract multiple subsamples from the training set; constructing a decision tree for each subsample, randomly selecting some features for evaluation at node splits; and finally determining the predicted category based on the voting results of the multiple decision trees.

[0038] In this application, the "feature_importance_" attribute of random forest was used to evaluate the influence weight of each feature (such as a specific tripeptide combination or physicochemical parameter) on inhibition specificity and identify key bioactivity determinants.

[0039] In this application, the SVM algorithm used achieves classification by finding the optimal hyperplane; mapping feature vectors into a high-dimensional space; maximizing the classification margin between positive and negative bacterial activity samples; and employing a radial basis kernel function (RBF) to handle nonlinear relationships. Specifically, a grid search is used to optimize the penalty parameter C and the kernel coefficient gamma to balance model complexity and classification error.

[0040] In this application, the decision tree algorithm modeling process includes: selecting the optimal splitting feature based on information gain or Gini coefficient; recursively constructing a tree structure, with each node representing a specific feature (e.g., "VGA tripeptide frequency > 0.1%"); and pre-pruning (max_depth = 5) to prevent overfitting. Finally, leaf nodes output the probability of positive / negative bacterial inhibition.

[0041] In this application, the conditional inference tree algorithm used includes: selecting split features based on statistical tests (such as chi-square test); controlling the p-value threshold (α=0.05) to ensure split significance; and automatically handling multicollinearity problems between features.

[0042] In this application, the implemented naive Bayes classifier includes: assuming that each feature condition is independent; calculating the prior probability (the activity ratio of two types of bacteria); and processing continuous features (such as isoelectric point) through Gaussian kernel density estimation.

[0043] S6. Determine the target prediction model based on the excellence evaluation results of multiple feature models after training.

[0044] Furthermore, the target prediction model is determined based on the excellence evaluation results of multiple feature models after training, including: using cross-validation to evaluate the set evaluation indicators of each machine learning algorithm under different feature models; based on the evaluation results, selecting the best performing model and feature combination as the target prediction model; setting evaluation indicators including accuracy, recall and Kappa coefficient.

[0045] It can be understood that in this embodiment, five machine learning algorithms, namely decision tree, random forest, support vector machine, conditional inference tree, and naive Bayes, are used to model the above-mentioned feature matrix respectively to obtain a frequency feature matrix model, a physicochemical property feature model, and a feature splicing model; the above-mentioned data set is divided into a training set (80%) and a test set (20%) to train the model; a 10-fold cross-validation is used to evaluate the performance of each model under different feature inputs (physicochemical properties, 3-mer frequency, splicing features); the evaluation indicators include accuracy, recall rate, Kappa coefficient, etc., and the model and feature combination with the best performance are selected as the target prediction model to ensure the reliability and effectiveness of the prediction results.

[0046] More specifically, 10-fold cross-validation involves randomly dividing the training set data into 10 subsets of roughly equal size. Each subset is used as a validation set, and the remaining 9 subsets are used as training sets for model training and validation. This process is repeated 10 times, ensuring that each subset serves as a validation set.

[0047] For example, suppose that after comparison, the random forest model, with the concatenated feature input, achieves 90% accuracy, 88% recall, and a Kappa coefficient of 0.85, outperforming all other models and feature combinations. In this case, the random forest model and the concatenated feature combination are selected as the target prediction model.

[0048] S7. Predict the antimicrobial peptide sequence to be tested based on the target prediction model to obtain its inhibition specificity classification results against Gram-positive bacteria and / or Gram-negative bacteria.

[0049] Furthermore, the feature values ​​of the sequence to be predicted are added to the composite feature matrix according to the same method, and then the trained random forest model is called to predict the sequence to be predicted. The model output is the specific inhibition result of the antimicrobial peptide against Gram-positive bacteria and / or Gram-negative bacteria, including: sequence ID, predicted specific classification result, such as Figure 3 shown.

[0050] Table 3 Output result list Sequence ID Prediction category Sequence1 Positive / negative Sequence 2 Positive / negative Sequence 3 Positive / negative Sequence 4 Positive / negative In one application scenario, the constructed optimal prediction model (the target prediction model obtained above) was used to predict the antibacterial activity of the candidate polypeptide sequences to be differentiated and identified against Gram-positive and Gram-negative bacteria. Nine candidate antimicrobial peptides that may be effective against Gram-positive bacteria and nine candidate antimicrobial peptides that may be effective against Gram-negative bacteria were randomly selected and chemically synthesized for experimental verification.

[0051] Escherichia coli and Staphylococcus aureus were selected as representative strains of Gram-negative and Gram-positive bacteria, respectively, for antibacterial activity experiments.

[0052] Specifically, the present application can divide each antimicrobial peptide into two groups. In one possible design, each representative antimicrobial peptide can be divided into a training group and a test group at a ratio of 1:1, that is, 50% as a training group and 50% as a test group. It should be noted that when performing grouping, the training group after grouping should be made to cover the above-mentioned two phenotypes, with high antibacterial activity against Gram-positive bacteria and high antibacterial activity against Gram-negative bacteria, and the test group after grouping should also cover the above-mentioned two phenotypes, with high antibacterial activity against Gram-positive bacteria and high antibacterial activity against Gram-negative bacteria.

[0053] The three matrix models, namely frequency characteristic matrix model, physical and chemical property characteristic model and characteristic splicing model, are named Model 1, Model 2 and Model 3 respectively.

[0054] Combine Figure 3 、 Figure 4 、 Figure 5As shown in the figure, the horizontal axes are CRT: Conditional Reasoning Tree Algorithm, Decision_tree: Decision Tree Algorithm, Naive_Bayes: Naive Bayes Algorithm, Random_Forest: Random Forest Algorithm, SVM: Support Vector Machine Algorithm, and the vertical axes are Accuracy, Kappa coefficient, Negative Recall, and Positive Recall. It can be seen from the figure that among the three matrix models, Model 2 has the best evaluation index, and the Random Forest Algorithm performs best among the five algorithms. Therefore, the Random Forest Algorithm of Model 2 is selected as the prediction model for the difference in antibacterial activity of antimicrobial peptides against Gram-positive bacteria and Gram-negative bacteria.

[0055] Combine Figure 6 As shown, the optimal prediction model was used to predict candidate antimicrobial peptide sequences. Nine candidate peptide sequences were randomly selected for in vitro validation. Antimicrobial activity was tested against representative strains of Escherichia coli and Staphylococcus aureus. The results showed that of the nine sequences predicted to be effective against Gram-positive bacteria, seven were correctly validated (with an accuracy rate of 77.78%). Of the nine sequences predicted to be effective against Gram-negative bacteria, five were correctly validated (with an accuracy rate of 55.56%), for an overall accuracy rate of 66.67%.

[0056] As can be appreciated, this application, by introducing a multi-source feature construction and multi-model integrated evaluation mechanism, breaks through the limitations of traditional reliance on single features or single algorithms, enabling accurate prediction of differences in antimicrobial peptide activity between Gram-positive and Gram-negative bacteria. Furthermore, this invention can not only be used to classify and predict the activity of known antimicrobial peptides, but also to discover potential new antimicrobial peptides with previously unidentified specific activities, thus possessing broad application prospects. Experimental verification has demonstrated that this method can effectively guide the synthesis and screening of candidate antimicrobial peptides, contributing to the development of new antimicrobial drugs.

[0057] In one embodiment, Figure 7 As shown, a prediction system for the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria is proposed, which is used to perform the prediction method in Example 1. The system includes: The data acquisition module 11 is used to obtain the antimicrobial peptide sequence to be processed that has antimicrobial activity against Gram-positive bacteria and / or Gram-negative bacteria; The fragment segmentation module 12 is used to segment the antimicrobial peptide sequence to be processed to obtain a multi-dimensional frequency feature matrix; A feature extraction module 13 is used to extract physical and chemical features of the antimicrobial peptide sequence to be processed to obtain a physical and chemical feature matrix; A feature splicing module 14 is used to splice the frequency feature matrix and the physical and chemical feature matrix to form a composite feature matrix; A model training module 15 is used to train multiple feature models constructed based on a machine learning algorithm, using a frequency feature matrix, a physical and chemical feature matrix, and a composite feature matrix as model inputs; A model determination module 16 is used to determine a target prediction model based on the excellence evaluation results of the multiple feature models after training; The prediction module 17 is used to predict the antimicrobial peptide sequence to be tested based on the target prediction model to obtain the specific classification results of its inhibition on Gram-positive bacteria and / or Gram-negative bacteria.

[0058] The specific limitations of the prediction system for the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria can be found in the limitations of the method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria described above and will not be further elaborated here. It should be noted that the modules in the above-described prediction system correspond to steps S1 to S7 of the above-described prediction method. The examples and application scenarios implemented by the various modules and corresponding steps are the same, but are not limited to those disclosed in Example 1 above.

[0059] In one embodiment, Figure 8 As shown, a computer device 21 is proposed. The computer device can be a terminal or a server. The computer device 21 includes a processor 23, a memory 24, and a network interface 25 connected via a system bus 22. The processor 23 of the computer device 21 is used to provide computing and control capabilities. The memory of the computer device 24 includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface 26 of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria is implemented.

[0060] In one embodiment, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by the processor 23 of the computer device 21 , the computer device 21 is enabled to perform the steps of the prediction method in any of the above embodiments.

[0061] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part.

[0062] The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated therein. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0063] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0064] In addition, the functional modules in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0065] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria, characterized in that: The method comprises: Obtaining antimicrobial peptide sequences to be processed that have antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria respectively; The antimicrobial peptide sequence to be processed is segmented to obtain a multi-dimensional frequency feature matrix; The physical and chemical characteristics of the antimicrobial peptide sequence to be processed are extracted to obtain a physical and chemical characteristic matrix; splicing the frequency characteristic matrix and the physical and chemical characteristic matrix to form a composite characteristic matrix; Multiple feature models built based on machine learning algorithms are trained using frequency feature matrices, physical and chemical feature matrices, and composite feature matrices as model inputs; Determine the target prediction model based on the excellence evaluation results of multiple feature models after training; The antimicrobial peptide sequences to be tested are predicted based on the target prediction model to obtain the specific classification results of their inhibition on Gram-positive bacteria and / or Gram-negative bacteria.

2. The method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to claim 1, characterized in that: The method of obtaining the antimicrobial peptide sequence to be processed that has antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria comprises: Obtaining antimicrobial peptide sequences that have antimicrobial activity against Gram-positive and Gram-negative bacteria; Adding a marker with inhibitory activity at a set position in the sequence; The repeated sequences in the antimicrobial peptide sequence are removed to obtain the antimicrobial peptide sequence to be processed.

3. The method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to claim 1, characterized in that: The antimicrobial peptide sequence to be processed is segmented to obtain a multi-dimensional frequency feature matrix, including: Sliding cuts are performed on the short peptide chain composed of amino acids in the antimicrobial peptide sequence to be processed, with three consecutive amino acids as a unit, to generate multiple tripeptide combinations; The occurrence frequencies of all tripeptide combinations in the antimicrobial peptide sequence to be processed are counted to construct a multidimensional frequency feature matrix, wherein each dimension in the frequency feature matrix corresponds to the frequency value of a tripeptide combination.

4. The method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to claim 1, wherein: The physicochemical characteristics of the antimicrobial peptide sequences to be processed were extracted, including the following extraction indicators: isoelectric point, sequence length, molecular weight, hydrophobicity, hydrophobic moment, disordered conformation tilt, hydrophilicity index and net charge.

5. The method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to claim 1, characterized in that: The multiple feature models constructed based on the machine learning algorithm are trained using the frequency feature matrix, the physical and chemical feature matrix, and the composite feature matrix as model inputs, respectively, and include: Construct frequency feature matrix models, physical and chemical property feature models, and feature splicing models based on machine learning algorithms; The frequency feature matrix, the physical and chemical feature matrix, and the composite feature matrix are respectively input into corresponding models as training sets for training; The machine learning algorithms used include decision trees, random forests, support vector machines, conditional inference trees, and naive Bayes.

6. The method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to claim 5, characterized in that: The step of determining the target prediction model based on the excellence evaluation results of the multiple feature models after training includes: Use cross-validation to evaluate the set evaluation indicators of each machine learning algorithm under different feature models; Based on the evaluation results, the best performing model and feature combination is selected as the target prediction model; The set evaluation indicators include precision, recall and Kappa coefficient.

7. The method for predicting the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria according to claim 5, characterized in that: The model output is the specific inhibition result of the antimicrobial peptide against Gram-positive bacteria and / or Gram-negative bacteria, including sequence ID and predicted specific classification result.

8. A prediction system for the inhibitory specificity of antimicrobial peptides against Gram-positive and Gram-negative bacteria, for executing the prediction method according to any one of claims 1 to 7, characterized in that: The system comprises: A data acquisition module is used to obtain antimicrobial peptide sequences to be processed that have antimicrobial activity against Gram-positive bacteria and Gram-negative bacteria respectively; The fragment segmentation module is used to segment the antimicrobial peptide sequence to be processed and obtain a multi-dimensional frequency feature matrix; A feature extraction module is used to extract the physical and chemical features of the antimicrobial peptide sequence to be processed and obtain a physical and chemical feature matrix; A feature splicing module, used for splicing the frequency feature matrix and the physical and chemical feature matrix to form a composite feature matrix; The model training module is used to train multiple feature models built based on machine learning algorithms, using frequency feature matrices, physical and chemical feature matrices, and composite feature matrices as model inputs. A model determination module is used to determine the target prediction model based on the excellence evaluation results of multiple feature models after training; The prediction module is used to predict the antimicrobial peptide sequence to be tested based on the target prediction model to obtain its inhibition specificity classification results on Gram-positive bacteria and / or Gram-negative bacteria.

9. A computer device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that When the instructions in the computer-readable storage medium are executed by a processor of a computer device, the computer device is enabled to perform the prediction method according to any one of claims 1 to 7.

Images