A protein mass spectrum data expansion method, system, computer device and medium

Abstract

The application belongs to the technical field of data recognition, and discloses a protein mass spectrum data expansion method, a protein mass spectrum data expansion system, computer equipment and a medium. The protein mass spectrum data expansion method comprises the following steps: for each mass spectrum data satisfying a sequence length greater than or equal to a sampled characteristic length, performing a with-return paired sampling on original mass spectrum data to expand a sample capacity, adding a Gaussian white noise module to each mass spectrum data after the sample capacity is expanded, so that the similarity between sample data exists; for each mass spectrum data satisfying a sequence length less than a sampled group number, adding a Gaussian white noise module to expand the sequence length of each original mass spectrum data, splicing each mass spectrum data after the sequence length is expanded with original mass spectrum data into new sample mass spectrum data, and then performing a with-return paired sampling to expand the sample capacity. The protein mass spectrum data expansion method has good expansion effect, and the obtained data is accurate.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of data recognition, and particularly relates to a protein mass spectrum data expansion method and system, a computer device and a medium. BACKGROUND

[0002] At present, mass spectrometry has been the dominant technology for peptide and protein identification and quantification for a long time. The typical identification strategy is performed in combination with a data-dependent acquisition (DDA) mode and a search engine. In a serial mode, data is collected in the DDA mode, and when a primary mass spectrum (MS1) is scanned, only the top k peptide ions are selected, which are fragmented to form a tandem mass spectrum (secondary mass spectrum MS2). The search engine matches the MS2 spectrum with the theoretical spectrum of the peptide to identify the peptide corresponding to the MS2 spectrum. However, because the top k peptide ions in intensity are randomly changed in repeated DDA experiments, the reproducibility of the peptides identified by the DDA method is poor. In order to overcome the limitations of the DDA mode, a data-independent acquisition (DIA) strategy appears. SWATH is a general DIA mode, the full name of which is Sequential Windowed Acquisition of All Theoretical Fragment Ions. The mass spectrometer fragments the peptide ions in each MS1 independent window and collects the signals of all fragment ions. Obviously, a large number of fragment signals of polypeptides are mixed in the corresponding secondary mass spectrum (MS2), that is, the sub-ion of the secondary mass spectrum is derived from multiple parent ions, and the fragment ions are also disturbed by the parent ions that are not fragmented. The DIA data is too complex, and it is extremely difficult to analyze directly.

[0003] Data augmentation is a simple technique to expand the training set based on generating synthetic image-label pairs. The idea is to transform images in a way that the label remains the same, or both the image and the label are defined as transformations, but since the augmentation methods such as random affine transformation, random elastic transformation and random contrast transformation have failed to significantly improve the segmentation performance of medical images, a new task-driven data augmentation method is used to learn limited labeled data, and a synthetic data generator can improve the optimization performance for the segmentation task.

[0004] Through the above analysis, the problems and defects of the prior art are that:

[0005] In the prior art, the protein mass spectrum data expansion effect is poor, and the obtained data is inaccurate. SUMMARY

[0006] In view of the problems in the prior art, the present application provides a protein mass spectrum data expansion method.

[0007] The present application is implemented by a protein mass spectrum data expansion method, which comprises the following steps:

[0008] For each mass spectrum data satisfying the sequence length greater than or equal to the sampled feature length, the original mass spectrum data is expanded in sample capacity by paired sampling with replacement, and Gaussian white noise is randomly added to each mass spectrum data after expansion of sample capacity, so that the similarity between sample data exists;

[0009] When the sequence length of each mass spectrum data is less than the group number of sampling, a Gaussian white noise module is added to expand the sequence length of each original mass spectrum data, and each mass spectrum data after expansion of sequence length is spliced with the original mass spectrum data to form new sample mass spectrum data, and then paired sampling with replacement is performed to expand the sample capacity;

[0010] The mass spectrum data with similarity and the new sample mass spectrum data spliced with the original mass spectrum data are represented by a sparse sequence and input into a classification network model.

[0011] Further, the Gaussian white noise module processes the mass spectrum data by Gaussian white noise to expand the M / Z and I sequence length under each original mass spectrum data.

[0012] Further, the processing of the mass spectrum data by the Gaussian white noise module includes:

[0013] Each protein mass spectrum data is defined as an original signal x (M / Z,I) (n), and the effective average intensity of the original signal is p signal As shown in the formula:

[0014]

[0015] The signal intensity of the noise is p noise , as shown in the formula:

[0016]

[0017]

[0018] First, the length of the original signal x(n) is calculated as len(x(n)), and then Gaussian white noise is used, the Gaussian white noise is a randn function, and the noise signal to be added is calculated by the formula:

[0019]

[0020] Finally, the output of the randomly added noisy sample mass spectrum data is shown in the formula, and each generated noisy sample mass spectrum data is spliced with the original mass spectrum data to obtain a new sample mass spectrum data pair of M / Z and I;

[0021] y (M / Z,I) =x (M / Z,I) (n)+Noise (M / Z,I) (5)。

[0022] Further, the protein mass spectrum data expansion method specifically comprises the following steps:

[0023] S1, set the sequence length of each original mass spectrum data to Length;

[0024] S2, when I>1, set the sequence length of each mass spectrum data to LengthA; when I≤1, set the sequence length of each mass spectrum data to LengthB; (Length=LengthA+LengthB);

[0025] S3, when I>1, if the LengthA sequence length of each mass spectrum data is greater than or equal to the group number N of each sampling, extract N groups of paired data of M / Z and I in a mass spectrum data with a sequence length of LengthA as a new sample mass spectrum data, and set the number of Bootstrap to m, to m sample mass spectrum data;

[0026] S4, if the LengthA length of each mass spectrum data is less than the group number N of each sampling, the sample mass spectrum data with a LengthA length is not allowed to be sampled;

[0027] S5, if the sequence length Length of each original mass spectrum data is greater than or equal to the group number N of each sampling, analyze the sequence length Length of one of the original mass spectrum data, and retain one mass spectrum data with a sequence length of LengthA;

[0028] S6, then add a Gaussian white noise module to one mass spectrum data with a sequence length of LengthB when I≤1, and the length of the mass spectrum data after passing through the Gaussian white noise module is LengthB*K, the value of K is 1 / 3 of the characteristic length, and the characteristic length at this time is N-LengthB;

[0029] S7, if the sequence length Length of each original mass spectrum data is less than the group number N of each sampling, then through the addition of a Gaussian white noise module, the length of each mass spectrum data is Length*K, the value of K is 1 / 3 of the characteristic length, and the characteristic length at this time is N;

[0030] S8, constructing sparse sequence representation of the oversampled data set;

[0031] S9, filling I data corresponding to each null sequence length in the size position of each mass spectrum data M / Z in turn;

[0032] S10, taking the sum of the sparse sequence data of the original I obtained by the original mass spectrum data through S9 and the sparse sequence data of the m I after expansion as a kind as the input end of the classification network model type.

[0033] Further, in step S3, Gaussian white noise is randomly added to the m sample mass spectrum data respectively, and m sample mass spectrum data after expansion of the original mass spectrum data is obtained, and the sequence length of each sample mass spectrum data is N.

[0034] Further, in step S6, the mass spectrum data after adding Gaussian white noise module is Bootstrap sampled m times, the sequence length of each sampled mass spectrum data is N-LengthB, and finally m sample mass spectrum data is spliced with the mass spectrum data with sequence length LengthA in turn, and m sample mass spectrum data after expansion of the original mass spectrum data is obtained, and the sequence length of each sample mass spectrum data is N.

[0035] Further, in step S7, one of the mass spectrum data is Bootstrap sampled m times, and the sequence length of each sampled sequence is N, so that one mass spectrum data is expanded to obtain m sample mass spectrum data.

[0036] In step S8, first, the maximum value of all mass spectrum data M / Z sequences is taken as the sparse sequence length of I, and the null sequence length is constructed.

[0037] In step S9, the value of I is filled in that M / Z is 1, and is increased to 1200 dimensions in turn with an interval of 1; if there is no I information in the M / Z position, it is filled with zero; if there are multiple I information in the same M / Z position, the sparse sequence representation of I mass spectrum data is obtained according to the principle of filling the maximum value of I according to the same M / Z position.

[0038] Another purpose of the present application is to provide a protein mass spectrum data expansion system comprising:

[0039] The sequence length setting module is used for setting the sequence length of each original mass spectrum data as Length.

[0040] A sequence length extraction module is configured to set the sequence length of each mass spectrum data as Length A when I>1, and set the sequence length of each mass spectrum data as Length B when I≤1, Length=Length A+Length B.

[0041] A sample mass spectrum data acquisition module is configured to, when I>1, if the Length A sequence length of each mass spectrum data is greater than or equal to the group number N of each sampling, extract N groups of paired data of M / Z and I in one mass spectrum data with the sequence length of Length A as a new sample mass spectrum data, so that the number of Bootstrap is set as m, and m sample mass spectrum data are obtained.

[0042] A sample mass spectrum data sampling judgment module is configured to, if the Length A sequence length of each mass spectrum data is less than the group number N of each sampling.

[0043] A mass spectrum data retention module is configured to, if the sequence length Length of each original mass spectrum data is greater than or equal to the group number N of each sampling.

[0044] A Gaussian white noise module is configured to, if the sequence length of each mass spectrum data is Length B when I≤1, the length of one mass spectrum data after Gaussian white noise is Length B*K.

[0045] If the sequence length Length of each original mass spectrum data is less than the group number N of each sampling, the length of each mass spectrum data is Length*K, one mass spectrum data is Bootstrap sampled m times, and the sequence length of each sampling is N, so that one mass spectrum data is expanded to obtain m sample mass spectrum data.

[0046] A null sequence length construction module is configured to construct a sparse sequence representation of an oversampled data set.

[0047] A sparse sequence representation mass spectrum data acquisition module is configured to sequentially fill the data of I in each null sequence length corresponding to the size position of each mass spectrum data M / Z.

[0048] A classification model acquisition module is configured to obtain the sum of the original I sparse sequence data and the expanded m I sparse sequence data as one category, so as to serve as the input end of the classification model.

[0049] Another object of the present application is to provide a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor to enable the processor to execute the protein mass spectrum data expansion method.

[0050] Another object of the present application is to provide a computer readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the protein mass spectrum data augmentation method.

[0051] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present application are analyzed from the following aspects:

[0052] In view of the technical problems existing in the prior art and the difficulty in solving the problems, in combination with the technical solutions to be protected by the present application and the results and data in the research and development process, the technical problems solved by the technical solutions of the present application are analyzed in detail and profoundly, and some creative technical effects brought about after the problems are solved are described as follows:

[0053] The mass-to-charge ratio (M / Z) and intensity (I) of the protein mass spectrum data are in one-to-one correspondence, and I is taken as a dividing point in combination with the sequence length of each original mass spectrum data (the number of M / Z and I groups under one mass spectrum data sample is taken as the sequence length).

[0054] First, the sequence length of each mass spectrum data is greater than or equal to the characteristic length (N groups) of sampling, and the original mass spectrum data is expanded in sample capacity by paired sampling with replacement, and Gaussian white noise is added to each mass spectrum data after the sample capacity is expanded to solve the similarity problem existing between the sample data. Second, the sequence length of each mass spectrum data is less than the number N of groups of sampling, and a Gaussian white noise module is added to expand the sequence length of each original mass spectrum data, and each mass spectrum data after the sequence length is expanded is spliced with the original mass spectrum data to form a new sample mass spectrum data, and then the sample capacity is expanded by paired sampling with replacement.

[0055] The mass spectrum data expanded by the above two ideas is represented by a sparse sequence and input into a classification network model. Therefore, the present application proposes a Flex-Bootstrap method. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is a protein mass spectrum data augmentation method flowchart provided by an embodiment of the present application;

[0057] Figure 2 is a Gaussian white noise module processing flowchart provided by an embodiment of the present application;

[0058] Figure 3 is a protein mass spectrum data augmentation system schematic diagram provided by an embodiment of the present application.

[0059] In the figure: 1, sequence length setting module; 2, sequence length extraction module; 3, sample mass spectrum data acquisition module; 4, sample mass spectrum data sampling judgment module; 5, mass spectrum data retention module; 6, Gaussian white noise module; 7, construct null sequence length module; 8, sparse sequence representation mass spectrum data acquisition module; 9, classification model acquisition module. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0061] I. Explanation of the embodiments. In order to enable those skilled in the art to fully understand how the present application is specifically implemented, this part is an explanation of the embodiments of the technical scheme of the claims.

[0062] Embodiment 1

[0063] As shown in the figure, the protein mass spectrum data expansion method provided by the embodiment of the present application comprises: Figure 1

[0064] S101, set the sequence length of each original mass spectrum data to Length.

[0065] S102, when I>1, set the sequence length of each mass spectrum data to LengthA; when I≤1, set the sequence length of each mass spectrum data to LengthB. (Length=LengthA+LengthB).

[0066] S103, when I>1, if the LengthA sequence length of each mass spectrum data is greater than or equal to the group number N of each sampling, extract N groups of M / Z and I paired data in a mass spectrum data with sequence length LengthA as a new sample mass spectrum data, so that the number of Bootstrap is set to m, and then m sample mass spectrum data is obtained. There is similarity between sample data, so Gaussian white noise is randomly added to m sample mass spectrum data, so that m sample mass spectrum data is obtained after expansion of the original mass spectrum data, and the sequence length of each sample mass spectrum data is N.

[0067] S104, if the length of each mass spectrum data LengthA is less than the group number N of each sampling, it means that the sample mass spectrum data with length LengthA is not allowed to be sampled. At this time, the sequence length of each original mass spectrum data is analyzed.

[0068] ​S105, if the sequence length Length of each original mass spectrum data is greater than or equal to the group number N of each sampling, the sequence length of one of the original mass spectrum data is Length, and one mass spectrum data with the sequence length LengthA is reserved.

[0069] S106, then a Gaussian white noise module is added to one mass spectrum data with the sequence length LengthB when I≤1, and the length of one mass spectrum data after the Gaussian white noise module is LengthB*K (the value of K is 1 / 3 of the characteristic length, and the characteristic length is N-LengthB at this time). The one mass spectrum data after the random Gaussian white noise module is Bootstrap sampled m times, and the sequence length of each sampling is N-LengthB. Finally, the m sample mass spectrum data are sequentially spliced with the reserved one mass spectrum data with the sequence length LengthA, and thus m sample mass spectrum data after expansion of one original mass spectrum data are obtained, and the sequence length of each sample mass spectrum data is N.

[0070] S107, if the sequence length Length of each original mass spectrum data is less than the group number N of each sampling, the length of each mass spectrum data is Length*K (the value of K is 1 / 3 of the characteristic length, and the characteristic length is N at this time) by adding a Gaussian white noise module, one mass spectrum data is Bootstrap sampled m times, and the sequence length of each sampling is N. Thus, one mass spectrum data is expanded to obtain m sample mass spectrum data.

[0071] S108, the construction of sparse sequence representation of the oversampled data set, first taking the maximum value of all mass spectrum data M / Z sequences plus a certain number as the sparse sequence length of I, and constructing the null sequence length in this way.

[0072] S109, filling the data of I in each null sequence length corresponding to the size position of each mass spectrum data M / Z in turn. The value of I filling is to start from 1 with an interval of 1 and increase to 1200 dimensions in turn. If there is no I information at the M / Z position, it is filled with zero. If there are multiple I information at the same M / Z position (due to oversampling), the sparse sequence representation mass spectrum data of I is obtained according to the principle of filling the maximum value of I with the same M / Z bit.

[0073] S110, the sum of the original I sparse sequence data obtained by S109 and the m I sparse sequence data after expansion is a kind, which is used as the input end of the classification model.

[0074] Example 2

[0075] Based on the protein mass spectrum data expansion method provided in Example 1 of the present application,Figure 2 As shown in the formula, the embodiment of the present application provides a Gaussian white noise processing procedure for mass spectrum data by using a Gaussian white noise module, and expands the M / Z and I sequence length under each original mass spectrum data.

[0076] Specifically, the following steps are included:

[0077] In the above procedure, each protein mass spectrum data is defined as original signal x (M / Z,I) (n), and the effective average intensity of the original signal is p signal As shown in the formula:

[0078]

[0079] As shown in the formula of signal-to-noise ratio (SNR), the signal intensity of the noise is p noise , and the formula is as shown:

[0080]

[0081]

[0082] First, the length of the original signal x(n) is calculated as len(x(n)), and then the Gaussian white noise, i.e., the randn function, is used. Thus, the noise signal to be added is shown in the formula:

[0083]

[0084] Finally, the output is the randomly added noisy sample mass spectrum data as shown in the formula, and each generated noisy sample mass spectrum data is correspondingly spliced with each original mass spectrum data to obtain new sample mass spectrum data in pairs of M / Z and I.

[0085] y (M / Z,I) =x (M / Z,I) (n)+Noise (M / Z,I) (5).

[0086] Embodiment 3

[0087] Based on the protein mass spectrum data expansion method provided in Embodiment 1 of the present application, as shown in the formula, the embodiment of the present application provides a protein mass spectrum data expansion system, which includes: Figure 3

[0088] A sequence length setting module 1 is used to set the sequence length of each original mass spectrum data as Length.

[0089] ​Sequence length extraction module 2, for setting the sequence length of each mass spectrum data to LengthA when I>1; setting the sequence length of each mass spectrum data to LengthB when I≤1. (Length=LengthA+LengthB)

[0090] Sample mass spectrum data acquisition module 3, for when I>1, if the LengthA sequence length of each mass spectrum data is greater than or equal to the group number N of each sampling, extracting N groups of M / Z and I paired data in a mass spectrum data with sequence length LengthA as a new sample mass spectrum data, so that the number of Bootstrap is set to m, thus obtaining m sample mass spectrum data. It is analyzed that there is similarity between sample data, so Gaussian white noise is added to m sample mass spectrum data respectively, so m sample mass spectrum data after expansion of the original mass spectrum data is obtained and the sequence length of each sample mass spectrum data is N.

[0091] Sample mass spectrum data sampling judgment module 4, for if the LengthA length of each mass spectrum data is less than the group number N of each sampling, it is indicated that the sample mass spectrum data with LengthA length is not allowed to be sampled. At this time, the sequence length of each original mass spectrum data is analyzed.

[0092] Mass spectrum data retention module 5, for if the sequence length Length of each original mass spectrum data is greater than or equal to the group number N of each sampling, the sequence length Length of one of the original mass spectrum data is analyzed, and one mass spectrum data with sequence length LengthA is retained.

[0093] Gaussian white noise module 6, for one mass spectrum data with sequence length LengthB when I≤1 is added to the Gaussian white noise module, and the length of one mass spectrum data after the Gaussian white noise module is LengthB*K (the value of K is 1 / 3 of the characteristic length, and the characteristic length at this time is N-LengthB). One mass spectrum data after random addition of the Gaussian white noise module is sampled m times by Bootstrap, and the sequence length of each sampling mass spectrum data is N-LengthB. Finally, m sample mass spectrum data are sequentially spliced with one mass spectrum data with sequence length LengthA retained, thus obtaining m sample mass spectrum data after expansion of the original mass spectrum data and the sequence length of each sample mass spectrum data is N.

[0094] If the sequence length Length of each original mass spectrum data is less than the group number N of each sampling, the length of each mass spectrum data is Length*K (K is 1 / 3 of the characteristic length, and the characteristic length is N at this time), one mass spectrum data is Bootstrap sampled m times, and the sequence length of each sampling is N, and m sample mass spectrum data is obtained by expanding one mass spectrum data.

[0095] The null sequence length module 7 is constructed for constructing the sparse sequence representation of the oversampled data set, and the maximum value of all mass spectrum data M / Z sequences is taken as the sparse sequence length of I, and the null sequence length is constructed by adding a certain value.

[0096] The sparse sequence representation mass spectrum data acquisition module 8 is used for sequentially filling the data of I at each null sequence length position of each mass spectrum data M / Z, and the value of I is filled starting from 1 to 1200 dimensions with an interval of 1. If there is no I information at the M / Z position, it is filled with zero. If there are multiple I information at the same M / Z position (due to oversampling), the sparse sequence representation mass spectrum data of I is obtained according to the principle of filling the maximum value of I at the same M / Z position.

[0097] The classification model acquisition module 9 takes the sum of the original I sparse sequence data and the expanded m I sparse sequence data as a category, and takes this as the input end of the classification model.

[0098] It should be noted that the embodiments of the present application can be realized by hardware, software or a combination of software and hardware. The hardware part can be realized by using special logic; the software part can be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or a specially designed hardware. Those skilled in the art can understand that the above-mentioned devices and methods can be realized by using computer executable instructions and / or included in processor control code, such as provided on a carrier medium, such as a magnetic disk, CD or DVD-ROM, a programmable memory, such as a read-only memory (firmware) or a data carrier, such as an optical or electronic signal carrier. The devices of the present application and their modules can be realized by hardware circuit, such as ultra-large-scale integrated circuit or gate array, semiconductor, such as logic chip, transistor, etc., or programmable hardware device, such as field programmable gate array, programmable logic device, etc., can also be realized by software executed by various types of processors, and can also be realized by a combination of the above-mentioned hardware circuit and software, such as firmware.

[0099] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any modification, equivalent replacement and improvement within the technical range disclosed by the present application and within the spirit and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method of protein mass spectrometry data augmentation, the method comprising: The protein mass spectrum data expansion method comprises: For each mass spectrum data satisfying the sequence length greater than or equal to the sampled feature length, the original mass spectrum data is expanded in a pair of back-to-back sampling, Gaussian white noise is randomly added to each mass spectrum data after the sample capacity is expanded, and similarity exists between the sample data; When the sequence length of each mass spectrum data is less than the group number of sampling, a Gaussian white noise module is added to expand the sequence length of each original mass spectrum data, each mass spectrum data after the sequence length is expanded is spliced with the original mass spectrum data to form new sample mass spectrum data, and the sample capacity is expanded by back-to-back pair sampling; The mass spectrum data with similarity, the new sample mass spectrum data and the original mass spectrum data are spliced and represented by a sparse sequence, and input into a classification network model; The Gaussian white noise module processes the mass spectrum data by Gaussian white noise to expand the mass-to-charge ratio M / Z and intensity I sequence length of each original mass spectrum data; The Gaussian white noise module processes the mass spectrum data by Gaussian white noise, which comprises: Each protein mass spectrometry data is defined as a primary signal x (M / Z,I) (n), the effective average intensity of the primary signal is p signal As shown in equation (1): The signal strength of the noise is p noise as shown in equation (3): First, the length of the original signal x(n) is calculated as len(x(n)), and then a Gaussian white noise is used, the Gaussian white noise is a randn function, and the noise signal to be added is shown in formula (4): Finally, the randomly added noisy sample mass spectrum data is output as shown in formula (5), and each noisy sample mass spectrum data generated is spliced with each original mass spectrum data to obtain new sample mass spectrum data in pairs of M / Z and I; y (M / Z,I) = x (M / Z,I) (n) + Noise (M / Z,I) (5).

2. The method of claim 1, wherein the protein mass spectrometry data augmentation is performed by a computer system. The protein mass spectrum data expansion method specifically comprises the following steps: S1, the sequence length of each original mass spectrum data is set as Length; S2, when I>1, the sequence length of each mass spectrum data is set as LengthA; when I≤1, the sequence length of each mass spectrum data is set as LengthB; Length=LengthA+LengthB; S3, when I>1, if the LengthA sequence length of each mass spectrum data is greater than or equal to the group number N of each sampling, N groups of paired data of M / Z and I in a mass spectrum data with a sequence length of LengthA are extracted as a new sample mass spectrum data, the number of Bootstrap is set as m, and m sample mass spectrum data are obtained; S4, if the LengthA of each mass spectrum data is less than the group number N of each sampling, the sample mass spectrum data with a length of LengthA is not allowed to be sampled; S5, if the sequence length Length of each original mass spectrum data is greater than or equal to the group number N of each sampling, the sequence length Length of one of the original mass spectrum data is analyzed, and one mass spectrum data with a sequence length of LengthA is reserved; S6, then a Gaussian white noise module is added to one mass spectrum data with a sequence length of LengthB when I≤1, the length of one mass spectrum data after the Gaussian white noise module is LengthB*K, and the value of K is 1 / 3 of the feature length, and the feature length at this time is N-LengthB; S7, if the sequence length Length of each original mass spectrum data is less than the group number N of each sampling, the length of each mass spectrum data is Length*K by adding a Gaussian white noise module, K is 1 / 3 of the characteristic length, and the characteristic length is N at this time; S8, constructing sparse sequence representation of the over-sampled data set; S9, sequentially filling I data in each null sequence length on the size position of each mass spectrum data M / Z; S10, summing the original I sparse sequence data obtained by S9 and the sparse sequence data of the expanded m I as a kind, as the input end of the classification network model type.

3. The method of claim 2, wherein the protein mass spectrometry data augmentation is performed by, In step S3, m sample mass spectrum data are obtained by adding Gaussian white noise to each sample mass spectrum data.

4. The method of claim 2, wherein the protein mass spectrometry data augmentation is performed by a neural network. 5 In step S6, one mass spectrum data after adding Gaussian white noise module is Bootstrap sampled m times, the sequence length of each sampling mass spectrum data is N-LengthB, and finally m sample mass spectrum data are sequentially spliced with one mass spectrum data with sequence length LengthA, to obtain m sample mass spectrum data after expanding the original mass spectrum data, and the sequence length of each sample mass spectrum data is N.

5. The protein mass spectrometry data augmentation method as described in claim 2, characterized in that, In step S7, one mass spectrum data is Bootstrap sampled m times and the sequence length of each sampling is N, so that m sample mass spectrum data are obtained by expanding one mass spectrum data; In step S8, the maximum value of all mass spectrum data M / Z sequences is taken as the sparse sequence length of I, and the null sequence length is constructed. In step S9, the value of I is filled in M / Z as 1, which is sequentially increased to 1200 dimensions; if there is no I information in the M / Z position, it is filled with zero; if there is more than one I information in the same M / Z position, the sparse sequence representation of I is obtained according to the principle of filling the maximum value of I according to the same M / Z bit.

6. A protein mass spectrum data augmentation system for implementing the protein mass spectrum data augmentation method according to any one of claims 1 to 5, characterized by The protein mass spectrum data expansion system comprises: A sequence length setting module for setting the sequence length of each original mass spectrum data as Length; A sequence length extraction module for setting the sequence length of each mass spectrum data as LengthA when I>1, and setting the sequence length of each mass spectrum data as LengthB when I≤1, Length=LengthA+LengthB; A sample mass spectrum data acquisition module for, when I>1, if the LengthA sequence length of each mass spectrum data is greater than or equal to the group number N of each sampling, extracting N groups of paired data of M / Z and I in one mass spectrum data with sequence length LengthA as a new sample mass spectrum data, so that the Bootstrap number is set as m, and m sample mass spectrum data are obtained; A sample mass spectrum data sampling judgment module for, if the LengthA length of each mass spectrum data is less than the group number N of each sampling; A mass spectrum data retention module is configured to retain each original mass spectrum data if the sequence length Length of each original mass spectrum data is greater than or equal to the number of groups N sampled each time; A Gaussian white noise module is configured to retain a mass spectrum data with a sequence length of LengthB*K after Gaussian white noise processing of a mass spectrum data with a sequence length of LengthB when I≤1; If the sequence length Length of each original mass spectrum data is less than the number of groups N sampled each time, the length of each mass spectrum data is Length*K, one of the mass spectrum data is Bootstrap sampled m times and the sequence length of each sampling is N, and then a mass spectrum data is obtained by expansion to obtain m sample mass spectrum data; A null sequence length construction module is configured to construct sparse sequence representation of the oversampled data set; A sparse sequence representation mass spectrum data acquisition module is configured to sequentially fill I data in each null sequence length on the size position of each mass spectrum data M / Z. A classification model acquisition module is configured to take the sum of the original I sparse sequence data and the m I sparse sequence data obtained after expansion as a category, and take the category as the input end of the classification model.

7. A computer device, comprising: The computer device includes a memory and a processor, the memory stores a computer program, and the computer program is executed by the processor to make the processor execute the protein mass spectrum data expansion method in any one of claims 1-5. 8.A computer readable storage medium storing a computer program, the computer program being executed by a processor to make the processor execute the protein mass spectrum data expansion method in any one of claims 1-5.

Images