Method, system and equipment for predicting 4mC point location in escherichia coli based on deep learning and medium

Abstract

The invention provides a method, a system, equipment and a medium for predicting 4mC point locations in escherichia coli based on deep learning, and belongs to the technical field of 4mC point location prediction.The method comprises the following steps that DNA sequence data of escherichia coli genomes is collected, and the collected data is preprocessed to construct a training data set and an independent test data set; performing feature coding on the DNA sequence data in the data set to generate feature vectors; constructing a prediction model based on the one-dimensional convolutional neural network, training the prediction model by using the training data set, optimizing model parameters through a 10-fold cross validation mode, and evaluating the performance of the prediction model through the independent test data set after the training is completed; and inputting the DNA sequence data of the escherichia coli genome to be detected into the prediction model passing the performance evaluation to obtain a prediction result of the 4mC point location in the DNA sequence. According to the method, prediction of the 4mC point location in the escherichia coli is achieved, the prediction accuracy of the model is high, and the generalization ability is high.

Description

Background Art

[0002] DNA methylation, especially 4-methylcytosine, also known as 4mC, plays a vital role in organisms, and is involved in regulating multiple key processes such as DNA replication, transcriptional regulation, and gene expression. Therefore, in genomic research, accurately identifying 4mC sites in the genome is of great significance for understanding the genetic function and disease mechanism of 4mC sites.

[0003] Currently, 4mC sites can be identified through a variety of experimental techniques, such as mass spectrometry, simplified representative sulfite sequences, and single-molecule real-time sequencing, but these technologies have problems such as high cost and low efficiency when processing large-scale sequencing data. A new generation of prediction models based on computational methods have begun to be used in identifying 4mC sites. For example, models such as iDNA4mC, iEC4mC-SVM, DNA4mc-LIP, and Meta-4mCpred have been proposed and applied to the prediction of 4mC sites in various species. However, most of these models are based on traditional machine learning algorithms, and their performance is limited by the complexity of feature extraction and the generalization ability of the model. In particular, a deep learning model 4mCCNN has emerged and is used to predict 4mC sites in various species, but this model is not accurate in predicting 4mC sites in specific species, especially Escherichia coli.

[0004] This is a shortcoming of the prior art. Therefore, in view of the above-mentioned defects in the prior art, it is very necessary to provide a method, system, equipment and medium for predicting 4mC sites in Escherichia coli based on deep learning. Summary of the invention

[0005] The above-mentioned 4mC site prediction in the prior art is crucial to organisms. The experimental technology for identifying 4mC sites is costly and inefficient. The new generation of prediction models is limited by the complexity of feature extraction and has low prediction accuracy. The present invention provides a method, system, device and medium for predicting 4mC sites in Escherichia coli based on deep learning to solve the above-mentioned technical problems.

[0006] In a first aspect, the present invention provides a method for predicting 4mC sites in Escherichia coli based on deep learning, comprising the following steps:

[0007] S1. Collect DNA sequence data of the Escherichia coli genome, preprocess the collected data to construct a training data set and an independent test data set;

[0008] S2. Feature encoding the DNA sequence data in the training data set and the independent test data set to generate a feature vector of a preset dimension;

[0009] S3. Build a prediction model based on a one-dimensional convolutional neural network, train the prediction model using feature vectors in the training dataset, optimize the model parameters through 10-fold cross validation, and evaluate the performance of the prediction model through an independent test dataset after training.

[0010] S4. Input the DNA sequence data of the Escherichia coli genome to be tested into the prediction model that has passed the performance evaluation to obtain the prediction results of the 4mC sites in the DNA sequence.

[0011] Furthermore, the specific steps of step S1 are as follows:

[0012] S11. Collect DNA sequence data of the Escherichia coli genome based on the iDNA4mC prediction tool, and annotate the DNA sequence data with 4mC sites and non-4mC sites;

[0013] S12. Analyze the collected DNA sequence data, calculate the similarity of each DNA sequence, delete one of the two DNA sequences whose similarity is higher than a preset sequence identity threshold, and complete data preprocessing;

[0014] S13. Preset a length P, select N groups of DNA sequence fragments containing 4mC sites as positive samples according to the length P from the DNA sequence data that has completed data preprocessing, and select N groups of DNA sequences that do not contain 4mC sites as negative samples according to the length P, to obtain a training data set;

[0015] S14. From the DNA sequence data that have completed data preprocessing, M groups of DNA sequence fragments containing 4mC sites are screened out according to the length P as positive samples, and M groups of DNA sequences that do not contain 4mC sites are screened out according to the length P as negative samples, to obtain an independent test data set, where the ratio of N to M is a preset value.

[0016] Furthermore, the specific steps of step S2 are as follows:

[0017] S21. Use the continuous bag-of-words model in the Word2Vec word embedding tool to convert the DNA sequence of each training sample in the training data set into a feature vector of a preset dimension w;

[0018] S22. Use the continuous bag-of-words model in the Word2Vec word embedding tool to convert the DNA sequence of each test sample in the independent test dataset into a feature vector of a preset dimension w.

[0019] Furthermore, the specific steps of step S21 are as follows:

[0020] S211. Use the continuous bag-of-words model of the Word2Vec word embedding tool as the encoding model;

[0021] S212. Set the DNA sequence of the training sample with a length of P as the input of the encoding model, set the dimension of the feature vector of the encoding model to w, and set the feature vector composed of each nucleotide in the DNA sequence as the output of the encoding model;

[0022] S213. Input the training sample into the set encoding model to train the context similarity of the DNA sequence, and obtain the training sample as a feature vector of w dimensions;

[0023] The specific steps of step S22 are as follows:

[0024] S221. Use the continuous bag-of-words model of the Word2Vec word embedding tool as the encoding model;

[0025] S222. Set the DNA sequence of the test sample with a length of P as the input of the encoding model, set the dimension of the feature vector of the encoding model to w, and set the feature vector composed of each nucleotide in the DNA sequence as the output of the encoding model;

[0026] S223. Input the test sample into the set encoding model to train the context similarity of the DNA sequence, and obtain the test sample as a feature vector of dimension w.

[0027] Furthermore, the specific steps of step S3 are as follows:

[0028] S31. construct a prediction model based on a one-dimensional convolutional neural network, wherein the prediction model includes several convolutional layers, several pooling layers and at least one fully connected layer, and initialize the parameters of the prediction model;

[0029] S32. randomly divide the training data set into a training subset and a test subset, use the training subset to train the prediction model, extract the hierarchical features, and use the test subset to evaluate the classification accuracy of the prediction model, and then adjust the parameters of the prediction model according to the evaluation results;

[0030] S33. Evaluate the performance of the trained prediction model using an independent test data set;

[0031] If the performance evaluation is passed, proceed to step S4;

[0032] If the performance evaluation fails, the prediction model training is performed again and the process returns to step S21.

[0033] Furthermore, the specific steps of step S32 are as follows:

[0034] S321. Randomly divide the training data set into 10 subsets and locate 1 subset;

[0035] S322. The located subset is used as a test subset, and the remaining 9 subsets are selected as training subsets;

[0036] S323. Train the prediction model using the training subset to extract hierarchical features;

[0037] S324. Evaluate the classification accuracy of the prediction model using the test subset and record the evaluation results;

[0038] S325. Determine whether the subset is located;

[0039] If yes, go to step S326;

[0040] If not, locate the next subset and return to step S322;

[0041] S326. Calculate the average of the 10 evaluation results as the final evaluation result;

[0042] If the final evaluation result meets the requirements, proceed to step S33;

[0043] If the final result does not meet the requirements, go to step S327;

[0044] S327. Adjust the parameters of the prediction model and return to step S21;

[0045] The specific steps of step S33 are as follows:

[0046] S331. Input the DNA sequence in the independent test data set into the trained prediction model to obtain the model prediction result;

[0047] S332. Compare the actual annotation results of the DNA sequences in the independent test data set with the model prediction results, and calculate the 4mC site recognition accuracy TP, the 4mC site recognition error rate FP, the non-4mC site recognition accuracy TN, and the non-4mC site recognition error rate FN;

[0048] S333. Calculate the sensitivity, specificity, accuracy and Matthews correlation coefficient of the prediction model according to the 4mC point recognition accuracy rate TP and the 4mC point recognition error rate FP;

[0049] S334. Constructing a working characteristic curve of the pre-model to characterize the efficiency of the prediction model, and calculating the classification efficiency area of ​​the working characteristic curve;

[0050] S335. Evaluate the prediction model using the prediction model sensitivity, specificity, accuracy, Matthews correlation coefficient index, and classification performance area of ​​the operating characteristic curve;

[0051] If the evaluation is passed, proceed to step S4;

[0052] If the evaluation fails, return to step S21.

[0053] Furthermore, the specific steps of step S4 are as follows:

[0054] S41. Obtain DNA sequence data of the Escherichia coli genome to be tested and input into the prediction model;

[0055] S42. The prediction model encodes, advances features and classifies the DNA sequence data of the Escherichia coli genome to be tested, and then outputs the prediction results to determine the 4mC site of the DNA sequence.

[0056] In a second aspect, the present invention provides a 4mC site prediction system in Escherichia coli based on deep learning, comprising:

[0057] A data set construction module is used to collect DNA sequence data of the Escherichia coli genome, and pre-process the collected data to construct a training data set and an independent test data set;

[0058] A feature vector generation module is used to perform feature encoding on the DNA sequence data in the training data set and the independent test data set to generate a feature vector of a preset dimension;

[0059] The model training module is used to build a prediction model based on a one-dimensional convolutional neural network, train the prediction model using feature vectors in the training data set, optimize the model parameters through a 10-fold cross-validation method, and evaluate the performance of the prediction model through an independent test data set after the training is completed;

[0060] The 4mC site prediction module is used to input the DNA sequence data of the Escherichia coli genome to be tested into the prediction model that has passed the performance evaluation to obtain the prediction result of the 4mC site in the DNA sequence.

[0061] In a third aspect, the present invention provides an electronic device, including a processor and a memory;

[0062] The memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the electronic device executes the method described in the first aspect above.

[0063] In a fourth aspect, the present invention provides a storage medium,

[0064] The storage medium stores instructions, which, when executed on a computer, enable the computer to execute the method described in the first aspect.

[0065] The beneficial effects of the present invention are:

[0066] The method, system, device and medium for predicting 4mC sites in Escherichia coli based on deep learning provided by the present invention realize the prediction of 4mC sites in Escherichia coli through a deep learning model based on a neural network, and improve the prediction accuracy and generalization ability of the model through a 10-fold cross-validation method during the model training process. It can be seen that compared with the prior art, the present invention has outstanding substantive characteristics and significant progress, and the beneficial effects of its implementation are also obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0068] Figure 1 It is a schematic diagram of the process of the method for predicting 4mC sites in Escherichia coli based on deep learning of the present invention.

[0069] Figure 2 It is a schematic diagram of the 4mC site prediction system in Escherichia coli based on deep learning of the present invention.

[0070] Figure 3 It is a schematic diagram of the prediction model structure constructed based on the one-dimensional convolutional neural network of the present invention.

[0071] Figure 4 It is a schematic diagram of the ROC curve of the 10-fold cross validation of the present invention. DETAILED DESCRIPTION

[0072] The deep learning-based 4mC site prediction method in Escherichia coli can be applied to one or more devices, wherein the device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), digital processors (DSP), embedded devices, etc.

[0073] An electronic device may be any electronic product that can interact with a user, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), an interactive network television (IPTV), etc. The network in which the electronic device is located includes but is not limited to the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), etc.

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0075] See also Figure 1 The figure shows a specific implementation of a method for predicting 4mC sites in Escherichia coli based on deep learning, comprising the following steps:

[0076] S1. Collect DNA sequence data of the Escherichia coli genome, preprocess the collected data to construct a training data set and an independent test data set;

[0077] S2. Feature encoding the DNA sequence data in the training data set and the independent test data set to generate a feature vector of a preset dimension;

[0078] S3. Build a prediction model based on a one-dimensional convolutional neural network, train the prediction model using feature vectors in the training dataset, optimize the model parameters through 10-fold cross validation, and evaluate the performance of the prediction model through an independent test dataset after training.

[0079] S4. Input the DNA sequence data of the Escherichia coli genome to be tested into the prediction model that has passed the performance evaluation to obtain the prediction results of the 4mC sites in the DNA sequence.

[0080] In this embodiment, the specific steps of step S1 are as follows:

[0081] S11. Collect DNA sequence data of the Escherichia coli genome based on the iDNA4mC prediction tool, and annotate the DNA sequence data with 4mC sites and non-4mC sites;

[0082] S12. Analyze the collected DNA sequence data, calculate the similarity of each DNA sequence, delete one of the two DNA sequences whose similarity is higher than a preset sequence identity threshold, and complete data preprocessing; for example, the sequence identity threshold can be set to 60%;

[0083] S13. Preset the length P, and select N groups of DNA sequence fragments containing 4mC sites as positive samples according to the length P from the DNA sequence data that has completed data preprocessing, and select N groups of DNA sequences that do not contain 4mC sites as negative samples according to the length P, to obtain a training data set; the length P can be set to 41bp, bp is the abbreviation of base pair, that is, base pair, which represents the number of bases, and 41bp actually means that the DNA sequence consists of 41 bases; N can be set to 270; that is, a benchmark data set of 270 positive samples and 270 negative samples with a length of 41bp is finally selected for training and testing the model;

[0084] S14. From the DNA sequence data after data preprocessing, M groups of DNA sequence fragments containing 4mC sites are screened out according to the length P as positive samples, and M groups of DNA sequences not containing 4mC sites are screened out according to the length P as negative samples, to obtain an independent test data set, wherein the ratio of N to M is a preset value; the length P can be set to 41 bp, and M can be set to 134; that is, 134 positive samples and 134 negative samples with a length of 41 bp are finally selected for independent testing;

[0085] The specific steps of step S2 are as follows:

[0086] S21. Use the continuous bag-of-words model in the Word2Vec word embedding tool to convert the DNA sequence of each training sample in the training data set into a feature vector of a preset dimension w; the specific steps of step S21 are as follows:

[0087] S211. Use the continuous bag-of-words model of the Word2Vec word embedding tool as the encoding model;

[0088] S212. Set the DNA sequence of the training sample with a length of P as the input of the encoding model, set the dimension of the feature vector of the encoding model to w, and set the feature vector composed of each nucleotide in the DNA sequence as the output of the encoding model;

[0089] S213. Input the training sample into the set encoding model to train the context similarity of the DNA sequence, and obtain the training sample as a feature vector of w dimensions;

[0090] S22. Use the continuous bag-of-words model in the Word2Vec word embedding tool to convert the DNA sequence of each test sample in the independent test data set into a feature vector of a preset dimension w; the specific steps of step S22 are as follows:

[0091] S221. Use the continuous bag-of-words model of the Word2Vec word embedding tool as the encoding model;

[0092] S222. Set the DNA sequence of the test sample with a length of P as the input of the encoding model, set the dimension of the feature vector of the encoding model to w, and set the feature vector composed of each nucleotide in the DNA sequence as the output of the encoding model;

[0093] S223. Input the test sample into the set encoding model to train the context similarity of the DNA sequence, and obtain the test sample as a feature vector of dimension w;

[0094] The structure of the continuous bag-of-words model shows the advantage of regularly organizing information in the data set. The continuous bag-of-words model is used to train the contextual similarity of DNA sequences. The dimension of the word2vec embedding feature vector is 200. The specific steps of step S3 are as follows:

[0095] S31. construct a prediction model based on a one-dimensional convolutional neural network, wherein the prediction model includes several convolutional layers, several pooling layers and at least one fully connected layer, and initialize the parameters of the prediction model;

[0096] In Python 3.5.4, Keras 2.3.1 and TensorFlow 2.1.0 are used to build the prediction model. The prediction model architecture and parameters are as follows: Figure 3 As shown, all experiments were conducted on a 1.6GHz Intel quad-core i5 with Windows operating system;

[0097] S32. Randomly divide the training data set into a training subset and a test subset, use the training subset to train the prediction model, extract the hierarchical features, and use the test subset to evaluate the classification accuracy of the prediction model, and then adjust the parameters of the prediction model according to the evaluation results; the specific steps of step S32 are as follows:

[0098] S321. Randomly divide the training data set into 10 subsets and locate 1 subset;

[0099] S322. The located subset is used as a test subset, and the remaining 9 subsets are selected as training subsets;

[0100] S323. Train the prediction model using the training subset to extract hierarchical features;

[0101] S324. Evaluate the classification accuracy of the prediction model using the test subset and record the evaluation results;

[0102] S325. Determine whether the subset is located;

[0103] If yes, go to step S326;

[0104] If not, locate the next subset and return to step S322;

[0105] S326. Calculate the average of the 10 evaluation results as the final evaluation result;

[0106] If the final evaluation result meets the requirements, proceed to step S33;

[0107] If the final result does not meet the requirements, go to step S327;

[0108] S327. Adjust the parameters of the prediction model and return to step S21;

[0109] S33. Evaluate the performance of the trained prediction model using an independent test data set;

[0110] If the performance evaluation is passed, proceed to step S4;

[0111] If the performance evaluation fails, the prediction model training is performed again and the process returns to step S21;

[0112] The specific steps of step S33 are as follows:

[0113] S331. Input the DNA sequence in the independent test data set into the trained prediction model to obtain the model prediction result;

[0114] S332. Compare the actual annotation results of the DNA sequences in the independent test data set with the model prediction results, and calculate the 4mC site recognition accuracy TP, the 4mC site recognition error rate FP, the non-4mC site recognition accuracy TN, and the non-4mC site recognition error rate FN;

[0115] S333. Calculate the prediction model sensitivity Sn, specificity Sp, accuracy Acc and Matthews correlation coefficient MCC according to the 4mC point recognition accuracy TP and the 4mC point recognition error rate FP;

[0116] Among them, the calculation formulas of the prediction model sensitivity Sn, specificity Sp, accuracy Acc and Matthews correlation coefficient index MCC are as follows:

[0117]

[0118]

[0119] S334. Constructing a pre-model operating characteristic curve to characterize the efficiency of the prediction model, and calculating the classification efficiency area AUC of the operating characteristic ROC curve;

[0120] The classification efficiency area AUC is the area under the operating characteristic curve. For example, the AUC of a good classifier is 1, while the AUC of random performance is 0.5.

[0121] S335. Evaluate the prediction model using the prediction model sensitivity, specificity, accuracy, Matthews correlation coefficient index, and classification performance area of ​​the operating characteristic curve;

[0122] If the evaluation is passed, proceed to step S4;

[0123] If the evaluation fails, return to step S21;

[0124] The DNA sequences of the independent test set were simultaneously input into the existing model 4mCCNN to calculate the sensitivity, specificity, accuracy, Matthews correlation coefficient index, and classification efficiency area of ​​the operating characteristic curve;

[0125] For example, the calculated AUC of the trained prediction model is 0.989, which is about 2.9% higher than the existing model 4mCCNN;

[0126] On the same independent test data set, i.e., 134 4mCs and 134 non-4mCs, the effectiveness of the prediction model of the present application and the above-mentioned existing models were measured; among them, the MCC, Sn, Sp, Acc and AUC of the existing model 4mCCNN were 0.653, 0.786, 0.766, 0.818 and 0.914 respectively; the MCC, Sn, Sp, Acc and AUC of the Projected Model of the present application were 0.670, 0.876, 0.773, 0.861 and 0.962 respectively; obviously, the model proposed by us is 4.8% higher than the existing model in AUC as shown in Table 1 below;

[0127] Table 1

[0128]

[0129] Among them, CV refers to the type of cross-validation, 10-fold is the 10-fold cross-validation method used in this application, and Ind-test is an independent test set;

[0130] It can be seen that the excellent efficiency of the prediction model of the present application is attributed to the use of a unique and precise encoding scheme and the selection of a suitable classifier;

[0131] The specific steps of step S4 are as follows:

[0132] S41. Obtain DNA sequence data of the Escherichia coli genome to be tested and input into the prediction model;

[0133] S42. The prediction model encodes, advances features and classifies the DNA sequence data of the Escherichia coli genome to be tested, and then outputs the prediction results to determine the 4mC site of the DNA sequence.

[0134] It should be understood that the order of execution of the steps in the above embodiment does not mean that the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0135] The following is an embodiment of the system for predicting the 4mC position in Escherichia coli based on deep learning provided in an embodiment of the present disclosure. The device and the method for predicting the 4mC position in Escherichia coli based on deep learning in the above-mentioned embodiments belong to the same inventive concept. For details not fully described in the embodiment of the system for predicting the 4mC position in Escherichia coli based on deep learning, please refer to the embodiment of the method for predicting the 4mC position in Escherichia coli based on deep learning.

[0136] like Figure 2 As shown, the system includes:

[0137] A data set construction module is used to collect DNA sequence data of the Escherichia coli genome, and pre-process the collected data to construct a training data set and an independent test data set;

[0138] A feature vector generation module is used to perform feature encoding on the DNA sequence data in the training data set and the independent test data set to generate a feature vector of a preset dimension;

[0139] The model training module is used to build a prediction model based on a one-dimensional convolutional neural network, train the prediction model using feature vectors in the training data set, optimize the model parameters through a 10-fold cross-validation method, and evaluate the performance of the prediction model through an independent test data set after the training is completed;

[0140] The 4mC site prediction module is used to input the DNA sequence data of the Escherichia coli genome to be tested into the prediction model that has passed the performance evaluation to obtain the prediction result of the 4mC site in the DNA sequence.

[0141] The component comparison method of the electric energy collection device for realizing carbon emission detection provided by the present invention is a combination of the modules and algorithm steps of each example described in the embodiments disclosed in this article, and can be implemented by electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. 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 to exceed the scope of the present invention.

[0142] Those skilled in the art will appreciate that the various aspects of the deep learning-based 4mC site prediction method in Escherichia coli provided by the present invention can be implemented as a system, method or program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which can be collectively referred to as "circuit", "module" or "system" here.

[0143] In the non-transitory computer-readable storage medium storing the 4mC site prediction method in Escherichia coli based on deep learning, it can be but not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0144] Although the present invention has been described in detail with reference to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.

Claims

1. A method for predicting 4mC sites in Escherichia coli based on deep learning, characterized in that: The steps include: S1. Collect DNA sequence data of the Escherichia coli genome, preprocess the collected data to construct a training data set and an independent test data set; S2. Feature encoding the DNA sequence data in the training data set and the independent test data set to generate a feature vector of a preset dimension; S3. Build a prediction model based on a one-dimensional convolutional neural network, train the prediction model using feature vectors in the training dataset, optimize the model parameters through 10-fold cross validation, and evaluate the performance of the prediction model through an independent test dataset after training. S4. Input the DNA sequence data of the Escherichia coli genome to be tested into the prediction model that has passed the performance evaluation to obtain the prediction results of the 4mC sites in the DNA sequence.

2. The method for predicting 4mC sites in Escherichia coli based on deep learning according to claim 1, characterized in that: The specific steps of step S1 are as follows: S11. Collect DNA sequence data of the Escherichia coli genome based on the iDNA4mC prediction tool, and annotate the DNA sequence data with 4mC sites and non-4mC sites; S12. Analyze the collected DNA sequence data, calculate the similarity of each DNA sequence, delete one of the two DNA sequences whose similarity is higher than a preset sequence identity threshold, and complete data preprocessing; S13. Preset a length P, select N groups of DNA sequence fragments containing 4mC sites as positive samples according to the length P from the DNA sequence data that has completed data preprocessing, and select N groups of DNA sequences that do not contain 4mC sites as negative samples according to the length P, to obtain a training data set; S14. From the DNA sequence data that have completed data preprocessing, M groups of DNA sequence fragments containing 4mC sites are screened out according to the length P as positive samples, and M groups of DNA sequences that do not contain 4mC sites are screened out according to the length P as negative samples, to obtain an independent test data set, where the ratio of N to M is a preset value.

3. The method for predicting 4mC sites in Escherichia coli based on deep learning according to claim 2, characterized in that: The specific steps of step S2 are as follows: S21. Use the continuous bag-of-words model in the Word2Vec word embedding tool to convert the DNA sequence of each training sample in the training data set into a feature vector of a preset dimension w; S22. Use the continuous bag-of-words model in the Word2Vec word embedding tool to convert the DNA sequence of each test sample in the independent test dataset into a feature vector of a preset dimension w.

4. The method for predicting 4mC sites in Escherichia coli based on deep learning according to claim 3, characterized in that: The specific steps of step S21 are as follows: S211. Use the continuous bag-of-words model of the Word2Vec word embedding tool as the encoding model; S212. Set the DNA sequence of the training sample with a length of P as the input of the encoding model, set the dimension of the feature vector of the encoding model to w, and set the feature vector composed of each nucleotide in the DNA sequence as the output of the encoding model; S213. Input the training sample into the set encoding model to train the context similarity of the DNA sequence, and obtain the training sample as a feature vector of w dimensions; The specific steps of step S22 are as follows: S221. Use the continuous bag-of-words model of the Word2Vec word embedding tool as the encoding model; S222. Set the DNA sequence of the test sample with a length of P as the input of the encoding model, set the dimension of the feature vector of the encoding model to w, and set the feature vector composed of each nucleotide in the DNA sequence as the output of the encoding model; S223. Input the test sample into the set encoding model to train the context similarity of the DNA sequence, and obtain the test sample as a feature vector of dimension w.

5. The method for predicting 4mC sites in Escherichia coli based on deep learning according to claim 3, characterized in that: The specific steps of step S3 are as follows: S31. construct a prediction model based on a one-dimensional convolutional neural network, wherein the prediction model includes several convolutional layers, several pooling layers and at least one fully connected layer, and initialize the parameters of the prediction model; S32. randomly divide the training data set into a training subset and a test subset, use the training subset to train the prediction model, extract the hierarchical features, and use the test subset to evaluate the classification accuracy of the prediction model, and then adjust the parameters of the prediction model according to the evaluation results; S33. Evaluate the performance of the trained prediction model using an independent test data set; If the performance evaluation is passed, proceed to step S4; If the performance evaluation fails, the prediction model training is performed again and the process returns to step S21.

6. The method for predicting 4mC sites in Escherichia coli based on deep learning according to claim 5, characterized in that: The specific steps of step S32 are as follows: S321. Randomly divide the training data set into 10 subsets and locate 1 subset; S322. The located subset is used as a test subset, and the remaining 9 subsets are selected as training subsets; S323. Train the prediction model using the training subset to extract hierarchical features; S324. Evaluate the classification accuracy of the prediction model using the test subset and record the evaluation results; S325. Determine whether the subset is located; If yes, go to step S326; If not, locate the next subset and return to step S322; S326. Calculate the average of the 10 evaluation results as the final evaluation result; If the final evaluation result meets the requirements, proceed to step S33; If the final result does not meet the requirements, go to step S327; S327. Adjust the parameters of the prediction model and return to step S21; The specific steps of step S33 are as follows: S331. Input the DNA sequence in the independent test data set into the trained prediction model to obtain the model prediction result; S332. Compare the actual annotation results of the DNA sequences in the independent test data set with the model prediction results, and calculate the 4mC site recognition accuracy TP, the 4mC site recognition error rate FP, the non-4mC site recognition accuracy TN, and the non-4mC site recognition error rate FN; S333. Calculate the sensitivity, specificity, accuracy and Matthews correlation coefficient of the prediction model according to the 4mC point recognition accuracy rate TP and the 4mC point recognition error rate FP; S334. Constructing a working characteristic curve of the pre-model to characterize the efficiency of the prediction model, and calculating the classification efficiency area of ​​the working characteristic curve; S335. Evaluate the prediction model using the prediction model sensitivity, specificity, accuracy, Matthews correlation coefficient index, and classification performance area of ​​the operating characteristic curve; If the evaluation is passed, proceed to step S4; If the evaluation fails, return to step S21.

7. The method for predicting 4mC sites in Escherichia coli based on deep learning according to claim 5, characterized in that: The specific steps of step S4 are as follows: S41. Obtain DNA sequence data of the Escherichia coli genome to be tested and input into the prediction model; S42. The prediction model encodes, advances features and classifies the DNA sequence data of the Escherichia coli genome to be tested, and then outputs the prediction results to determine the 4mC site of the DNA sequence.

8. A 4mC site prediction system in Escherichia coli based on deep learning, characterized in that: include: A data set construction module is used to collect DNA sequence data of the Escherichia coli genome, and pre-process the collected data to construct a training data set and an independent test data set; A feature vector generation module is used to perform feature encoding on the DNA sequence data in the training data set and the independent test data set to generate a feature vector of a preset dimension; The model training module is used to build a prediction model based on a one-dimensional convolutional neural network, train the prediction model using feature vectors in the training data set, optimize the model parameters through a 10-fold cross-validation method, and evaluate the performance of the prediction model through an independent test data set after the training is completed; The 4mC site prediction module is used to input the DNA sequence data of the Escherichia coli genome to be tested into the prediction model that has passed the performance evaluation to obtain the prediction result of the 4mC site in the DNA sequence.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store a computer program, and the processor is used to call and run the computer program from the memory, so that the electronic device executes the method described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores instructions, which, when executed on a computer, enable the computer to execute the method described in any one of claims 1 to 7.

Images