RNA sequence coding potential prediction method and system

Abstract

The present invention provides a method and system for predicting the coding potential of an RNA sequence. The method comprises: selecting a coding RNA with an ORF length of less than 303 nt from a training sample set as a sample to be enhanced; performing data enhancement processing on the sample to be enhanced to obtain 36 different pseudo RNA sequences; combining the 36 different pseudo RNA sequences with the original training data to obtain 36 training sets; training 36 basic classifiers through the 36 training sets to obtain 36 basic classification models; connecting the output results of the 36 basic classification models in series into a 36-dimensional feature vector; selecting a result feature vector from the 36-dimensional feature vector; inputting the result feature vector and the label into the final classifier for training to obtain an RNA sequence coding potential prediction model, and predicting the RNA sequence coding potential through the RNA sequence coding potential prediction model. Through the present invention, the accuracy of RNA sequence coding potential prediction is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of bioinformatics, and in particular to a method and system for predicting RNA sequence coding potential. Background Art

[0002] High-throughput sequencing technologies have generated a large number of transcripts, which are combinations of DNA transcription products, including coding RNA and non-coding RNA (ncRNA). ncRNA refers to RNA that does not encode proteins and was once thought to be unrelated to gene expression. Later, it was recognized that ncRNA plays a key role in gene expression regulation and disease pathogenesis. Estimating the coding potential of transcripts, that is, distinguishing between coding RNA and ncRNA, is crucial for downstream biological function analysis.

[0003] Researchers at home and abroad have proposed many computational methods for predicting the coding potential of RNA sequences. They use different features and different classifiers to build prediction models. The successful application of these methods shows that feature engineering and classifier engineering have been widely studied in the problem of predicting the coding potential of RNA sequences.

[0004] How to further improve the accuracy of predicting the coding potential of RNA sequences is an urgent problem to be solved. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a method and system for predicting the coding potential of RNA sequences.

[0006] In a first aspect, the present invention provides a method for predicting the coding potential of an RNA sequence, the method comprising:

[0007] Select coding RNA with ORF length less than 303 nt from the training sample set as samples to be enhanced;

[0008] Data augmentation was performed on the samples to be augmented, and 36 different pseudo RNA sequences were obtained;

[0009] Combining 36 different pseudo-RNA sequences with the original training data, 36 training sets were obtained;

[0010] 36 basic classifiers are trained using 36 training sets to obtain 36 basic classification models;

[0011] The output results of the 36 basic classification models are concatenated into a 36-dimensional feature vector;

[0012] Select the resulting feature vector from the 36-dimensional feature vector;

[0013] The resulting feature vector and label are input into the final classifier for training to obtain the RNA sequence coding potential prediction model, and the RNA sequence coding potential is predicted by the RNA sequence coding potential prediction model.

[0014] Optionally, data augmentation processing includes:

[0015] Base-based data augmentation processing and subsequence-based data augmentation processing.

[0016] Optionally, the step of performing data enhancement processing on the samples to be enhanced to obtain 36 different types of pseudo RNA sequences includes:

[0017] The characters in the enhanced samples are replaced, inserted, and deleted with a probability of p. There are 6 values ​​of p, and 18 different pseudo RNA sequences are obtained.

[0018] The subsequence in the sample to be enhanced is reversed, copied, and deleted. The length of the subsequence is Len. The position of the first character of the subsequence in the sample to be enhanced is a random number between [0, L-Len]. There are 6 values ​​of Len, and 18 different pseudo RNA sequences are obtained.

[0019] Optionally, the basic classifier and the final classifier are both eXtreme Gradient Boosting extreme gradient boosting algorithms.

[0020] Optionally, after obtaining the RNA sequence coding potential prediction model, the method further comprises:

[0021] The test samples are predicted using the RNA sequence coding potential prediction model;

[0022] Compare the predictions with the true labels of the test samples;

[0023] Based on the comparison results, the evaluation results of the RNA sequence coding potential prediction model were obtained.

[0024] In a second aspect, the present invention further provides a RNA sequence coding potential prediction system, the RNA sequence coding potential prediction system comprising:

[0025] The first selection module is used to select coding RNA with an ORF length less than 303 nt from the training sample set as a sample to be enhanced;

[0026] The data enhancement module is used to perform data enhancement processing on the samples to be enhanced and obtain 36 different types of pseudo RNA sequences;

[0027] The combining module is used to combine 36 different pseudo RNA sequences with the original training data to obtain 36 training sets;

[0028] The first training module is used to train 36 basic classifiers through 36 training sets to obtain 36 basic classification models;

[0029] A concatenation module is used to concatenate the output results of 36 basic classification models into a 36-dimensional feature vector;

[0030] The second selection module is used to select a result feature vector from the 36-dimensional feature vector;

[0031] The second training module is used to input the result feature vector and the label into the final classifier for training to obtain the RNA sequence coding potential prediction model, and to predict the RNA sequence coding potential through the RNA sequence coding potential prediction model.

[0032] Optionally, the data enhancement processing includes:

[0033] Base-based data augmentation processing and subsequence-based data augmentation processing.

[0034] Optionally, the data enhancement module is used to:

[0035] The characters in the enhanced samples are replaced, inserted, and deleted with a probability of p. There are 6 values ​​of p, and 18 different pseudo RNA sequences are obtained.

[0036] The subsequence in the sample to be enhanced is reversed, copied, and deleted. The length of the subsequence is Len. The position of the first character of the subsequence in the sample to be enhanced is a random number between [0, L-Len]. There are 6 values ​​of Len, and 18 different pseudo RNA sequences are obtained.

[0037] Optionally, the basic classifier and the final classifier are both eXtreme Gradient Boosting extreme gradient boosting algorithms.

[0038] Optionally, the RNA sequence coding potential prediction system further includes an evaluation module, which is used to:

[0039] The test samples are predicted using the RNA sequence coding potential prediction model;

[0040] Compare the predictions with the true labels of the test samples;

[0041] Based on the comparison results, the evaluation results of the RNA sequence coding potential prediction model were obtained.

[0042] In the present invention, a coding RNA with an ORF length less than 303nt is selected from a training sample set as a sample to be enhanced; data enhancement processing is performed on the sample to be enhanced to obtain 36 different pseudo RNA sequences; the 36 different pseudo RNA sequences are combined with the original training data to obtain 36 training sets; 36 basic classifiers are trained by the 36 training sets to obtain 36 basic classification models; the output results of the 36 basic classification models are connected in series into a 36-dimensional feature vector; the result feature vector is selected from the 36-dimensional feature vector; the result feature vector and the label are input into the final classifier for training to obtain an RNA sequence coding potential prediction model, and the RNA sequence coding potential prediction is performed by the RNA sequence coding potential prediction model. Through the present invention, based on data enhancement processing, the generated pseudo RNA sequence is used for the training of the RNA sequence coding potential prediction model, so that the prediction performance of the RNA sequence coding potential prediction model finally obtained is effectively improved, thereby improving the accuracy of RNA sequence coding potential prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a schematic diagram of a process of an embodiment of a method for predicting RNA sequence coding potential of the present invention;

[0044] Figure 2 A schematic diagram of a scenario of an embodiment of a method for predicting RNA sequence coding potential of the present invention;

[0045] Figure 3 Schematic diagram of data enhancement processing;

[0046] Figure 4 Schematic diagram for evaluating the prediction model of RNA sequence coding potential;

[0047] Figure 5 Schematic diagram of the architecture of an embodiment of the RNA sequence coding potential prediction system of the present invention.

[0048] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0049] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0050] In a first aspect, an embodiment of the present invention provides a method for predicting the coding potential of an RNA sequence.

[0051] In one embodiment, referring to Figure 1 , Figure 1 Schematic diagram of a method for predicting RNA sequence coding potential according to an embodiment of the present invention. Figure 1 As shown, the RNA sequence coding potential prediction method includes:

[0052] Step S10, selecting coding RNA with an ORF length less than 303 nt from the training sample set as a sample to be enhanced;

[0053] Step S20, performing data enhancement processing on the samples to be enhanced to obtain 36 different types of pseudo RNA sequences;

[0054] Step S30, combining 36 different types of pseudo RNA sequences with the original training data to obtain 36 training sets;

[0055] Step S40, training 36 basic classifiers using 36 training sets to obtain 36 basic classification models;

[0056] Step S50, concatenating the output results of the 36 basic classification models into a 36-dimensional feature vector;

[0057] Step S60, selecting a result feature vector from the 36-dimensional feature vector;

[0058] Step S70, inputting the result feature vector and the label into the final classifier for training to obtain the RNA sequence coding potential prediction model, and performing RNA sequence coding potential prediction by using the RNA sequence coding potential prediction model.

[0059] In this embodiment, refer to Figure 2 , Figure 2 Schematic diagram of a scenario of an embodiment of the method for predicting RNA sequence coding potential of the present invention. Figure 2As shown in the figure, firstly, the coding RNA with ORF (Open Reading Frame, which is a nucleotide fragment consisting of three parts: the start codon, the stop codon and the nucleotide sequence between the two) less than 303nt in length is selected from the coding RNA of the training sample set as the sample to be enhanced (i.e., coding RNAs with sORF). Secondly, on this basis, different sequence variation methods are used to generate 36 different pseudo RNA sequences (i.e., pseudo sample sequences 1 to 36). These different types of pseudo RNA sequences are combined with the original training data to form 36 training sets (i.e., new training sets 1 to 36). These 36 training sets are used to train 36 basic classifiers to obtain 36 basic classification models. Specifically, the primary features of all RNA sequences in the 36 training sets are used to train 36 basic classifiers, and these primary features are sent to the trained basic classification model to obtain the output of 36 basic classifiers. Among them, the primary features of RNA sequences include: ORF length, ORF coverage, ORF completeness, Fickett score, Hexamer score, predicted peptide isoelectric point (pI), predicted peptide hydrophilicity (Gravy) Grand average, predicted peptide stability estimation and other 8 features, the global description feature CTD (composition, transition and distribution) includes 30 features such as nucleotide composition, nucleotide transition and nucleotide distribution, and all feature values ​​include these 38 features.

[0060] The results of these 36 basic classification models (i.e., labels 1 to 36) are concatenated and encoded into a 36-dimensional vector as a new feature vector for characterizing RNA sequences. After feature selection, the 36-dimensional feature vector of the training sample is subjected to the most important 5-dimensional feature vector (i.e., the result feature vector). The 5-dimensional feature vector of the training sample and its label are input into the final classifier for training to obtain the RNA sequence coding potential prediction model, which can be used to predict the RNA sequence coding potential. Among them, the feature selection method adopts the mRMR (minimum redundancy maximum relevance) technology.

[0061] In this embodiment, a coding RNA with an ORF length less than 303nt is selected from a training sample set as a sample to be enhanced; data enhancement processing is performed on the sample to be enhanced to obtain 36 different pseudo RNA sequences; the 36 different pseudo RNA sequences are combined with the original training data to obtain 36 training sets; 36 basic classifiers are trained by the 36 training sets to obtain 36 basic classification models; the output results of the 36 basic classification models are connected in series into a 36-dimensional feature vector; the result feature vector is selected from the 36-dimensional feature vector; the result feature vector and the label are input into the final classifier for training to obtain an RNA sequence coding potential prediction model, and the RNA sequence coding potential prediction is performed by the RNA sequence coding potential prediction model. Through this embodiment, based on data enhancement processing, the generated pseudo RNA sequence is used for the training of the RNA sequence coding potential prediction model, so that the prediction performance of the RNA sequence coding potential prediction model finally obtained is effectively improved, thereby improving the accuracy of RNA sequence coding potential prediction.

[0062] Furthermore, in one embodiment, the data enhancement process includes:

[0063] Base-based data augmentation processing and subsequence-based data augmentation processing.

[0064] In this embodiment, the base-based data enhancement processing and the subsequence-based data enhancement processing respectively obtain 18 different types of pseudo RNA sequences, and a total of 36 different types of pseudo RNA sequences are obtained.

[0065] Furthermore, in one embodiment, step S20 includes:

[0066] The characters in the enhanced samples are replaced, inserted, and deleted with a probability of p. There are 6 values ​​of p, and 18 different pseudo RNA sequences are obtained.

[0067] The subsequence in the sample to be enhanced is reversed, copied, and deleted. The length of the subsequence is Len. The position of the first character of the subsequence in the sample to be enhanced is a random number between [0, L-Len]. There are 6 values ​​of Len, and 18 different pseudo RNA sequences are obtained.

[0068] In this embodiment, R represents an RNA sequence of a sample to be enhanced, as follows:

[0069] R=N1…N i-1 N i …N j N j+1 …N L

[0070] Where 1≤i≤j≤L, L represents the length of the RNA sequence, each character in the RNA sequence is any one of A, U, G, C, and each character in the RNA sequence may be replaced by a new character with probability p, and the new character is also any one of A, U, G, C. p equals 0.001, which means that 1 out of every 1000 characters in the RNA sequence is replaced. The proportion of the changed characters to the characters in the entire RNA sequence is controlled by p.

[0071] Reference Figure 3 , Figure 3 This is a schematic diagram of data enhancement processing. Figure 3 As shown, Figure 3 (a) Replace the characters in R with probability p; Figure 3 (b) randomly insert a new character into the characters in R with probability p; Figure 3 (c) Randomly delete characters in R with probability p. p is an important parameter, and its value is any value in the set {0.001, 0.002, 0.003, 0.004, 0.005, 0.006}.

[0072] In this way, for the enhanced samples, 18 different types of pseudo RNA sequences can be obtained.

[0073] Further, Figure 3 (d) is a random reversal of a subsequence in the RNA sequence. The RNA sequence after reversal can be expressed as follows:

[0074] R new =N1…N i-1 N j …N i N j+1 …N L

[0075] The subsequence N j …N i is the subsequence N in the original RNA sequence i …N j reversal.

[0076] Figure 3 (e) A certain subsequence in the RNA sequence is replicated multiple times. The number of repetitions is a random number between [2, Max], and Max is set to 50. The RNA sequence after repetition can be expressed as follows:

[0077] R new =N1…N i-1 N i …N j ...N i …N j Nj+1 …N L

[0078] Figure 3 (f) is to randomly delete a subsequence in the RNA sequence. The RNA sequence after deletion can be expressed as follows:

[0079] R new =N1…N i-1 N j+1 …N L

[0080] In these three subsequence-based operations, the length of the subsequence is controlled by the parameter Len, and the position of the first character in the fragment in the original sequence is a random number between [0, L-Len]. Len is an important parameter, and its value is any value in the set {10, 20, 30, 40, 50, 60}. In this way, for the enhanced samples, 18 different pseudo RNA sequences can be obtained.

[0081] In summary, by performing data augmentation processing on the samples to be enhanced, 36 different types of pseudo RNA sequences can be obtained.

[0082] Furthermore, in one embodiment, the basic classifier and the final classifier are both eXtremeGradientBoosting extreme gradient boosting algorithms.

[0083] In this embodiment, the basic classifier and the final classifier used are both XGBoost. XGBoost is the abbreviation of ExtremeGradient Boosting. This algorithm is an improvement of gradientboosting decision tree. XGBoost has many advantages, such as: adding regularization terms to prevent overfitting, supporting parallelization, shrinkage technology, etc.

[0084] Furthermore, in one embodiment, after step S70, the method further includes:

[0085] The test sample is predicted by the RNA sequence coding potential prediction model; the prediction result is compared with the true label of the test sample; based on the comparison result, the evaluation result of the RNA sequence coding potential prediction model is obtained.

[0086] In this embodiment, continue to refer to Figure 2 After obtaining the RNA sequence coding potential prediction model, the test sample is input into the RNA sequence coding potential prediction model, and the prediction result of the RNA sequence coding potential prediction model for the input test sample is compared with the true label of the test sample; based on the comparison result, the evaluation result of the RNA sequence coding potential prediction model can be obtained.

[0087] In one embodiment, comparison is performed on four test sets, the names of which are: Human-testing, Human-sORF-testing, Mouse-testing, Mouse-sORF-testing, and the evaluation index used is MCC (Matthews Correlation Coefficient). Human-testing is data from humans after redundancy removal. Human-sORF-testing includes sequences with sORF length less than 303 nt selected from Human-testing. Mouse-testing is data from mice, and Mouse-sORF-testing includes sequences with sORF length less than 303 nt selected from Mouse-testing.

[0088] In order to evaluate the performance of the RNA sequence coding potential prediction model, the RNA sequence coding potential prediction model is preliminarily tested and compared on four test sets. MCC is a relatively comprehensive indicator. The larger the value of this indicator, the better the performance of the algorithm. Data augmentation technology is an important part of this method. The number of new samples N is an important parameter, which can affect the final performance of the model. The samples that need to be enhanced in the training set are coding RNAs with sORFs. The number of these samples is 449, so we set N to a multiple of 449 in the experiment, that is, N = 0, 898, 1796, ..., 12572. We use these enhanced training sets to establish different prediction models. These enhanced data sets have different numbers of new samples, and use different prediction models to evaluate the four test sets.

[0089] Reference Figure 4 , Figure 4 Schematic diagram of evaluating the RNA sequence coding potential prediction model. Figure 4 shows the impact of N on the MCC score of the RNA sequence coding potential prediction model on four test data sets. Figure 4 (a) It can be seen that for Human-testing and Mouse-testing, as N increases, the MCC value will have two peaks and then begin to decrease. Figure 4 In (b), as N increases, the MCC value of the model continues to increase and reaches a peak when N equals 7184. Figure 4 It is shown that the performance of the RNA sequence coding potential prediction model based on the enhanced training samples exceeds the performance of the model based on the original training samples (N=0), and the performance of the model reaches the best value when N is equal to 7184. In the following experiments, N is set to 7184.

[0090] Therefore, this embodiment can effectively improve the prediction performance of RNA sequence coding potential.

[0091] In a second aspect, an embodiment of the present invention also provides a system for predicting RNA sequence coding potential.

[0092] In one embodiment, referring to Figure 5 , Figure 5 FIG. 1 is a schematic diagram of the architecture of an embodiment of the RNA sequence coding potential prediction system of the present invention. Figure 5 As shown, the RNA sequence coding potential prediction system includes:

[0093] The first selection module 10 is used to select coding RNA with an ORF length less than 303 nt from the training sample set as a sample to be enhanced;

[0094] The data enhancement module 20 is used to perform data enhancement processing on the samples to be enhanced to obtain 36 different types of pseudo RNA sequences;

[0095] A combining module 30 is used to combine 36 different types of pseudo RNA sequences with the original training data to obtain 36 training sets;

[0096] A first training module 40 is used to train 36 basic classifiers using 36 training sets to obtain 36 basic classification models;

[0097] A concatenation module 50 is used to concatenate the output results of the 36 basic classification models into a 36-dimensional feature vector;

[0098] A second selection module 60 is used to select a result feature vector from the 36-dimensional feature vector;

[0099] The second training module 70 is used to input the result feature vector and the label into the final classifier for training to obtain the RNA sequence coding potential prediction model, and perform RNA sequence coding potential prediction through the RNA sequence coding potential prediction model.

[0100] Furthermore, in one embodiment, the data enhancement processing includes:

[0101] Base-based data augmentation processing and subsequence-based data augmentation processing.

[0102] Furthermore, in one embodiment, the data enhancement module 20 is used to:

[0103] The characters in the enhanced samples are replaced, inserted, and deleted with a probability of p. There are 6 values ​​of p, and 18 different pseudo RNA sequences are obtained.

[0104] The subsequence in the sample to be enhanced is reversed, copied, and deleted. The length of the subsequence is Len. The position of the first character of the subsequence in the sample to be enhanced is a random number between [0, L-Len]. There are 6 values ​​of Len, and 18 different pseudo RNA sequences are obtained.

[0105] Furthermore, in one embodiment, the basic classifier and the final classifier are both eXtremeGradientBoosting extreme gradient boosting algorithms.

[0106] Furthermore, in one embodiment, the RNA sequence coding potential prediction system further includes an evaluation module for:

[0107] The test samples are predicted using the RNA sequence coding potential prediction model;

[0108] Compare the predictions with the true labels of the test samples;

[0109] Based on the comparison results, the evaluation results of the RNA sequence coding potential prediction model were obtained.

[0110] Among them, the functional implementation of each module in the above-mentioned RNA sequence coding potential prediction system corresponds to the steps in the above-mentioned RNA sequence coding potential prediction method embodiment, and its functions and implementation processes are no longer repeated here.

[0111] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0112] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0113] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device to execute the methods described in each embodiment of the present invention.

[0114] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for predicting RNA sequence coding potential, characterized in that: The RNA sequence coding potential prediction method comprises: Select coding RNA with ORF length less than 303 nt from the training sample set as samples to be enhanced; Data augmentation was performed on the samples to be augmented, and 36 different pseudo RNA sequences were obtained; Combining 36 different pseudo-RNA sequences with the original training data, 36 training sets were obtained; 36 basic classifiers are trained using 36 training sets to obtain 36 basic classification models; The output results of the 36 basic classification models are concatenated into a 36-dimensional feature vector; Select the resulting feature vector from the 36-dimensional feature vector; The resulting feature vector and label are input into the final classifier for training to obtain an RNA sequence coding potential prediction model, and the RNA sequence coding potential is predicted by the RNA sequence coding potential prediction model; Data enhancement processing includes: Base-based data augmentation processing and subsequence-based data augmentation processing; The step of performing data enhancement processing on the samples to be enhanced to obtain 36 different types of pseudo RNA sequences includes: The characters in the enhanced samples are replaced, inserted, and deleted with a probability of p. There are 6 values ​​of p, and 18 different pseudo RNA sequences are obtained. The subsequence in the sample to be enhanced is reversed, copied, and deleted. The length of the subsequence is Len. The position of the first character of the subsequence in the sample to be enhanced is a random number between [0, L-Len]. There are 6 values ​​of Len, and 18 different pseudo RNA sequences are obtained.

2. The RNA sequence coding potential prediction method according to claim 1, wherein The basic classifier and the final classifier are both eXtreme Gradient Boosting extreme gradient boosting algorithms.

3. The RNA sequence coding potential prediction method according to claim 1, wherein After obtaining the RNA sequence coding potential prediction model, the method further includes: The test samples are predicted using the RNA sequence coding potential prediction model; Compare the predictions with the true labels of the test samples; Based on the comparison results, the evaluation results of the RNA sequence coding potential prediction model were obtained.

4. A RNA sequence coding potential prediction system, characterized in that: The RNA sequence coding potential prediction system comprises: The first selection module is used to select coding RNA with an ORF length less than 303 nt from the training sample set as a sample to be enhanced; The data enhancement module is used to perform data enhancement processing on the samples to be enhanced and obtain 36 different types of pseudo RNA sequences; The combining module is used to combine 36 different pseudo RNA sequences with the original training data to obtain 36 training sets; The first training module is used to train 36 basic classifiers through 36 training sets to obtain 36 basic classification models; A concatenation module is used to concatenate the output results of 36 basic classification models into a 36-dimensional feature vector; The second selection module is used to select a result feature vector from the 36-dimensional feature vector; The second training module is used to input the result feature vector and the label into the final classifier for training, so as to obtain the RNA sequence coding potential prediction model, and perform RNA sequence coding potential prediction through the RNA sequence coding potential prediction model; The data enhancement process includes: Base-based data augmentation processing and subsequence-based data augmentation processing; The data enhancement module is used to: The characters in the enhanced samples are replaced, inserted, and deleted with a probability of p. There are 6 values ​​of p, and 18 different pseudo RNA sequences are obtained. The subsequence in the sample to be enhanced is reversed, copied, and deleted. The length of the subsequence is Len. The position of the first character of the subsequence in the sample to be enhanced is a random number between [0, L-Len]. There are 6 values ​​of Len, and 18 different pseudo RNA sequences are obtained.

5. The RNA sequence coding potential prediction system according to claim 4, wherein: The basic classifier and the final classifier are both eXtreme Gradient Boosting extreme gradient boosting algorithms.

6. The RNA sequence coding potential prediction system according to claim 4, wherein: The RNA sequence coding potential prediction system also includes an evaluation module, which is used to: The test samples are predicted using the RNA sequence coding potential prediction model; Compare the predictions with the true labels of the test samples; Based on the comparison results, the evaluation results of the RNA sequence coding potential prediction model were obtained.

Images