Fingerprint spectrum generation method, device and system, and computer readable storage medium

Abstract

The present disclosure relates to a method, device and system for generating a fingerprint spectrum, and a computer readable storage medium, and relates to the technical field of image processing. The method for generating a fingerprint spectrum comprises: determining an intra-group feature of each spectrum of a plurality of spectra of a target type of lipid; determining an inter-group feature of the plurality of spectra; and generating a fingerprint spectrum of the target type of lipid according to the intra-group feature and the inter-group feature.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of image processing, and particularly relates to a method, device and system for generating a fingerprint spectrum, and a computer readable storage medium. BACKGROUND

[0002] Mass spectrometry is an analysis method widely used in the fields of environment, biology, medicine, etc. Mass spectrum of different substances can be obtained by mass spectrometry instrument. Substance identification is achieved by comparing mass spectrometry data with standard library and characteristic peak spectrum library information, which is an important application of mass spectrometry.

[0003] In the related art, a method of obtaining reference standard spectrum on a mass spectrometry instrument is used to construct a fingerprint spectrum library. SUMMARY

[0004] According to a first aspect of the present disclosure, a method for generating a fingerprint spectrum is provided, comprising: determining an intra-group feature of each spectrum of a plurality of spectra of a target category of lipids; determining an inter-group feature of the plurality of spectra; and generating a fingerprint spectrum of the target category of lipids according to the intra-group features and the inter-group feature.

[0005] In some embodiments, the determining the intra-group feature of each spectrum of the plurality of spectra of the target category of lipids comprises: extracting the intra-group feature from the each fingerprint spectrum by using a first activation layer and a linear layer of a fingerprint spectrum generation model.

[0006] In some embodiments, the determining the inter-group feature of the plurality of spectra comprises: determining the inter-group feature of the plurality of spectra according to the intra-group features of the plurality of spectra.

[0007] In some embodiments, the determining the inter-group feature of the plurality of spectra according to the intra-group features of the plurality of spectra comprises: determining the inter-group feature of the plurality of spectra according to the intra-group features of the plurality of spectra by using a multi-head attention layer of a fingerprint spectrum generation model.

[0008] In some embodiments, the generating the fingerprint spectrum of the target category of lipids according to the intra-group features and the inter-group feature comprises: generating a normalized feature according to the intra-group features and the inter-group feature by using a normalization layer of a fingerprint spectrum generation model; and generating the fingerprint spectrum of the target category of lipids according to the normalized feature.

[0009] In some embodiments, the generating the fingerprint spectrum of the target category of lipids according to the normalized feature comprises: generating an activation feature according to the normalized feature by using a second activation layer of a fingerprint spectrum generation model; and generating the fingerprint spectrum of the target category of lipids according to the activation feature.

[0010] In some embodiments, the generating the fingerprint spectrum of the target category of lipids according to the activation features comprises: utilizing an up-sampling layer of a fingerprint spectrum generation model to generate the fingerprint spectrum of the target category of lipids according to the activation features.

[0011] In some embodiments, the resolution of the plurality of spectra is a first resolution, and the resolution of the fingerprint spectrum is a second resolution, the second resolution being greater than the first resolution.

[0012] According to a second aspect of the present disclosure, there is provided a fingerprint spectrum generation apparatus, comprising: an intra-group feature determination module configured to determine an intra-group feature of each spectrum of a plurality of spectra of a target category of lipids; an inter-group feature determination module configured to determine an inter-group feature of the plurality of spectra; and a generation module configured to generate a fingerprint spectrum of the target category of lipids according to the intra-group feature and the inter-group feature.

[0013] According to a third aspect of the present disclosure, there is provided a fingerprint spectrum generation apparatus, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute a fingerprint spectrum generation method according to some embodiments of the present disclosure based on instructions stored in the memory.

[0014] According to a fourth aspect of the present disclosure, there is provided a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement a fingerprint spectrum generation method according to some embodiments of the present disclosure.

[0015] According to a fifth aspect of the present disclosure, there is provided a computer program product comprising computer program instructions which, when executed by a processor, implement a fingerprint spectrum generation method according to some embodiments of the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings, which constitute a part of this specification, illustrate embodiments of the present disclosure and serve to explain the principles of the present disclosure.

[0017] The present disclosure can be understood more readily by reference to the following detailed description, taken in connection with the accompanying drawings, and wherein:

[0018] Figure 1 a flowchart illustrating a fingerprint spectrum generation method according to some embodiments of the present disclosure;

[0019] Figure 2 a schematic diagram illustrating a model according to some embodiments of the present disclosure;

[0020] Figure 3 a schematic diagram illustrating a model according to some other embodiments of the present disclosure;

[0021] Figure 4 a block diagram illustrating a device for generating a fingerprint spectrum according to some embodiments of the present disclosure;

[0022] Figure 5 a block diagram illustrating a device for generating a fingerprint spectrum according to some embodiments of the present disclosure;

[0023] Figure 6 a block diagram illustrating a computer system for implementing some embodiments of the present disclosure;

[0024] Figure 7 a user interface of a generation system according to some embodiments of the present disclosure. DETAILED DESCRIPTION

[0025] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of the components and steps set forth in the embodiments, the numerical expressions, and the numerical values are not limiting to the scope of the present disclosure unless otherwise specifically stated.

[0026] It should be understood, of course, that the various embodiments of the disclosure are merely examples of implementations and are not limiting.

[0027] The following description of at least one example embodiment is merely illustrative in nature and is in no way limiting to the scope of the disclosure and its applications or uses.

[0028] Techniques, methods, and devices known to those of ordinary skill in the relevant art can not be discussed in detail herein. However, where appropriate, such techniques, methods, and devices can be viewed as part of the specification.

[0029] In all of the examples shown and discussed herein, any specific values should be interpreted as merely illustrative and not as a limitation of the exemplary embodiments. Thus, other examples of the exemplary embodiments can have different values.

[0030] It should be noted that like numbers and letters refer to like items throughout the drawings, and that, as such, no further discussion on such items is necessary.

[0031] In the related art, a library of fingerprint spectra is constructed by acquiring reference standard spectra on a mass spectrometry instrument. Due to the large number of lipid species, it is difficult and costly to synthesize reference standards, and only a small number of lipids have standard reference materials, so it is difficult to construct a comprehensive library of lipid molecular fingerprint spectra.

[0032] Lipids are a class of organic compounds composed of elements such as carbon, hydrogen, and oxygen, usually containing long-chain hydrocarbons (fatty acids) and connected with polar groups (hydroxyl, phosphate groups, etc.). The standard representation of lipids is composed of lipid subcategory, total carbon number: double bond number, and chain carbon number: double bond number, such as PC 36:3 (PC 18:1_18:2). Due to the high cost of synthesis and the large number of lipids, even the same subcategory with different side chain combinations can reach tens of millions of types, so there is a lack of comprehensive standards. Because a large number of lipids lack standard products, there is a lack of fingerprint spectrum. The fingerprint spectrum under specific conditions is usually obtained by mass spectrometry of lipid standards.

[0033] The method, device and system for generating a fingerprint spectrum, and the computer readable storage medium according to some embodiments of the present disclosure realize the generation of a fingerprint spectrum.

[0034] Figure 1 A flowchart of a method for generating a fingerprint spectrum according to some embodiments of the present disclosure is shown.

[0035] As shown in Figure 1 The method for generating a fingerprint spectrum includes steps S1-S3. In some embodiments, the method for generating a fingerprint spectrum is performed by a fingerprint spectrum generation model.

[0036] In step S1, the intra-group features of each spectrum of a plurality of spectra of a target type of lipids are determined; in step S2, the inter-group features of the plurality of spectra are determined; and in step S3, a fingerprint spectrum of the target type of lipids is generated according to the intra-group features and the inter-group features.

[0037] For example, if a fingerprint spectrum of a target type of lipids needs to be generated, the experimental spectra of samples of the type of lipids are obtained. Using a fingerprint spectrum generation model, the intra-group features within a single experimental spectrum and the inter-group features between a plurality of experimental spectra are learned, and a fingerprint spectrum of the target type of lipids is generated according to the intra-group features and the inter-group features.

[0038] According to some methods for generating a fingerprint spectrum of the present disclosure, the spectrum features are learned based on an artificial intelligence model, and the generation of a fingerprint spectrum is realized, thereby realizing the mass spectrometric identification of lipid substances.

[0039] The present disclosure realizes the generation of a fingerprint spectrum using the Kolmogorov-Arnold representation theorem. The Kolmogorov-Arnold representation theorem shows that a complex multivariate function can be represented as a combination of multiple univariate functions.

[0040] Lipids are lack of comprehensive standards due to high synthesis cost, numerous quantity, even the same sub-class with different side chain combinations, up to tens of millions of kinds. The present disclosure is based on deep learning, and reversely constructs the fingerprint spectrum. Based on Kolmogorov-Arnold representation theorem, the present disclosure regards the final fingerprint spectrum as a linear representation of multiple experimental spectra, and the principle can be expressed as follows, wherein, represents the final generated fingerprint spectrum, X represents multiple experimental spectra, W and B represent parameters, and n represents the number of experiments.

[0041] (1)

[0042] The present disclosure constructs a model for learning and generating a fingerprint spectrum under specific experimental conditions from an input spectrum, so as to improve the accuracy and coverage of identification.

[0043] First, the extraction of the intra-group feature is introduced.

[0044] In some embodiments, the intra-group feature of each spectrum of the multiple spectra of the target kind of lipids is determined, comprising: extracting the intra-group feature from each fingerprint spectrum by using the first activation layer and the linear layer of the fingerprint spectrum generation model.

[0045] For example, the fingerprint spectrum generation model simultaneously inputs multiple spectra, and realizes the extraction of the features of the spectra based on the Kolmogrov-Arnold theorem, so as to realize the intra-group feature learning. The network mainly learns the inherent features of the individual spectrum in the input spectrum, maps the secondary spectrum of the lipids to a fixed 512-row low-dimensional space, and cuts the data into eight matrices through the deflection of two Sigmoid activation functions (SiLU) and one rectified linear unit (ReLU) activation. The formula is as follows:

[0046]

[0047]

[0048] (2)

[0049] wherein, is the input matrix, is the weight matrix, is the bias function, is the output of the corresponding layer.

[0050] Through the first activation layer and the linear layer, the data is mapped to the latent space dimension, the features of the data are extracted, and the input problem of different dimensions is solved.

[0051] Next, the extraction of the inter-group feature is introduced.

[0052] In some embodiments, the inter-group features of the plurality of spectral graphs are determined according to the intra-group features of the plurality of spectral graphs. For example, the inter-group features of the plurality of spectral graphs are determined by analyzing and comparing the feature differences between different groups according to the intra-group features of the plurality of spectral graphs.

[0053] In some embodiments, the inter-group features of the plurality of spectral graphs are determined according to the intra-group features of the plurality of spectral graphs, including: utilizing a multi-head attention layer of the fingerprint spectrum generation model, determining the inter-group features of the plurality of spectral graphs according to the intra-group features of the plurality of spectral graphs.

[0054] The multi-head attention part of the fingerprint spectrum generation model, for example, includes 6 parallel encoders. The plurality of encoders performs inter-group feature learning in parallel, and learns the feature information between the plurality of spectral graphs through the global feature learning characteristics of the multi-head self-attention mechanism. This part contains an eight-head self-supervised module and a network layer. The eight matrices after segmentation respectively enter an eight-head self-supervised module, and the formula is as follows.

[0055]

[0056] (3)

[0057] wherein, , , , , by function splicing, the inter-group features .

[0058] In some embodiments, the fingerprint spectrum of the target type of lipid is generated according to the intra-group features and the inter-group features, including: utilizing a normalization layer of the fingerprint spectrum generation model, generating normalized features according to the intra-group features and the inter-group features; generating the fingerprint spectrum of the target type of lipid according to the normalized features.

[0059] For example, after completing the self-supervised learning, the feature matrix is subjected to a dropout function (Dropout) for overfitting control and a normalization layer (Norm) for output arrangement, and the formula is as follows.

[0060] (4)

[0061] In some embodiments, the fingerprint spectrum of the target type of lipid is generated according to the normalized features, including: utilizing a second activation layer of the fingerprint spectrum generation model, generating activation features according to the normalized features; generating the fingerprint spectrum of the target type of lipid according to the activation features.

[0062] For example, the result of the above process is subjected to a rectified linear unit (ReLU) for format arrangement, and the formula is as follows.

[0063] (5)

[0064] After the entire encoder processing is completed, the data is subjected to a preliminary resolution enhancement through an activation function (SiLU):

[0065] (6)

[0066] In some embodiments, the fingerprint spectrum of the target category of lipids is generated according to the activation features, including: using the up-sampling layer of the fingerprint spectrum generation model to generate the fingerprint spectrum of the target category of lipids according to the activation features.

[0067] After the foregoing processing, the model has learned the feature learning of multiple spectra, and can finally integrate and enhance the resolution of the data through up-sampling and down-sampling, as shown below.

[0068] (7)

[0069] In some embodiments, the resolution of the multiple spectra is a first resolution, and the resolution of the fingerprint spectrum is a second resolution, and the second resolution is greater than the first resolution. That is, through up-sampling, the high-resolution fingerprint spectrum is regenerated.

[0070] According to some embodiments of the present disclosure, the Kolmogorov-Arnold representation theorem is combined with the convolution network to extract the intra-data local features, and the multi-head attention is further used to extract the inter-data global features, thereby improving the feature extraction capability of the model and enabling better prediction effect in the analysis of complex data.

[0071] According to some embodiments of the present disclosure, a reverse omics model, i.e., a fingerprint spectrum generation model, is constructed. The fingerprint spectrum generation model realizes the extraction of fragment information from low-resolution fingerprint spectrum and the regeneration of high-resolution fingerprint spectrum by learning the intra-group and inter-group features of the input spectrum.

[0072] In some embodiments, the learning rate of the model is dynamically adjusted according to the different learning efficiencies of different data, thereby optimizing the training process of the model and significantly shortening the training time. At the same time, a convenient data batch processing function is provided, which is suitable for large-scale data processing and analysis.

[0073] Moreover, according to some embodiments of the present disclosure, the weights of the model can be pre-trained, and the model is robust and can be used for data of multiple platforms and is easy to migrate between platforms.

[0074] Through the trained model, high-precision fingerprint spectrum generation can be achieved even in the case of low input precision or poor instruments.

[0075] The fingerprint spectrum generated by the method of generating a fingerprint spectrum according to some embodiments of the present disclosure (referred to as a predicted spectrum in the table below) is compared with the experimental results of lipid markers sampled separately on a mass spectrometer instrument (referred to as an independently collected spectrum in the table below) for spectrum similarity comparison, and the results are as follows.

[0076] Table 1

[0077]

[0078] As can be seen from the above table, the reproducibility of the generated fingerprint spectrum is high.

[0079] The method of the present disclosure can be applied to process data generated by commonly used mass spectrometers on the market, such as Thermo Scientific Orbitrap Exploris 240 MS system, Thermo Scientific Orbitrap Exploris 120 MS system, Agilent 6546 Q-TOF MS system, Xevo G2-XS Q-TOF MS system, SCIEX ZenoTOF 7600 MS system, etc.

[0080] Figure 2 A schematic diagram of a model according to some embodiments of the present disclosure is shown.

[0081] As shown in Figure 2 , the model includes sub-models one, two, three, respectively for generating intra-group features, inter-group features, and fingerprint spectrum.

[0082] Figure 3 A schematic diagram of a model according to some embodiments of the present disclosure is shown.

[0083] Figure 3 In the above formula, w and b represent network parameters, and Relu represents a rectified linear unit. The spectrum data of multiple batches is input as the first network architecture (the uppermost input), which is processed through a two-layer network. The spectrum data is first mapped to a latent feature space, and then the relu activation function is used to obtain the features of the multiple batches of spectrum data. The obtained feature results are input as the input of the second network architecture (the middle input), which is processed through a multi-head self-attention structure to learn the features between the spectrum data and obtain the final output. The output results are input into the third network architecture (the lowermost input), which is processed through a two-layer network to upsample the feature information to a higher dimension, and then the dimensions are combined to finally generate and output a spectrum with higher accuracy.

[0084] Figure 4 A block diagram of a fingerprint spectrum generation device according to some embodiments of the present disclosure is shown.

[0085] As shown in Figure 4 , the fingerprint spectrum generation apparatus comprises an intra-group feature determination module 41, an inter-group feature determination module 42, and a generation module 43.

[0086] The intra-group feature determination module 41 is configured to determine the intra-group feature of each spectrum of the plurality of spectra of the target category of lipids, for example, by performing step S1 as shown in Figure 1 .

[0087] The inter-group feature determination module 42 is configured to determine the inter-group feature of the plurality of spectra, for example, by performing step S2 as shown in Figure 1 .

[0088] The generation module 43 is configured to generate the fingerprint spectrum of the target category of lipids according to the intra-group feature and the inter-group feature, for example, by performing step S3 as shown in Figure 1 .

[0089] In some embodiments, the intra-group feature determination module 41 is further configured to extract the intra-group feature from each fingerprint spectrum by using a first activation layer and a linear layer of a fingerprint spectrum generation model.

[0090] In some embodiments, the inter-group feature determination module 42 is further configured to determine the inter-group feature of the plurality of spectra according to the intra-group features of the plurality of spectra.

[0091] In some embodiments, the inter-group feature determination module 42 is further configured to determine the inter-group feature of the plurality of spectra according to the intra-group features of the plurality of spectra by using a multi-head attention layer of a fingerprint spectrum generation model.

[0092] In some embodiments, the generation module 43 is further configured to generate a normalized feature according to the intra-group feature and the inter-group feature by using a normalization layer of a fingerprint spectrum generation model; and generate the fingerprint spectrum of the target category of lipids according to the normalized feature.

[0093] In some embodiments, the generation module 43 is further configured to generate an activation feature according to the normalized feature by using a second activation layer of a fingerprint spectrum generation model; and generate the fingerprint spectrum of the target category of lipids according to the activation feature.

[0094] In some embodiments, the generation module 43 is further configured to generate the fingerprint spectrum of the target category of lipids according to the activation feature by using an up-sampling layer of a fingerprint spectrum generation model.

[0095] In some embodiments, the plurality of spectra has a first resolution, and the fingerprint spectrum has a second resolution, the second resolution being greater than the first resolution.

[0096] Figure 5A block diagram of a device for generating a fingerprint spectrum according to some embodiments of the present disclosure is shown.

[0097] As shown in Figure 5 the device for generating a fingerprint spectrum 5 includes a memory 51 and a processor 52 coupled to the memory 51. The memory 51 is configured to store instructions for performing a method for generating a fingerprint spectrum. The processor 52 is configured to perform the method for generating a fingerprint spectrum according to any of the embodiments of the present disclosure based on the instructions stored in the memory 51.

[0098] Figure 6 A block diagram of a computer system for implementing some embodiments of the present disclosure is shown.

[0099] As shown in Figure 6 the computer system 60 can be in the form of a general purpose computing device. The computer system 60 includes a memory 610, a processor 620 and a bus 600 connecting different system components.

[0100] The memory 610 can include, for example, system memory, non-volatile storage media and the like. The system memory, for example, stores an operating system, application programs, a Boot Loader and other programs and the like. The system memory can include volatile storage media such as random access memory (RAM) and / or cache memory. The non-volatile storage media, for example, stores instructions for performing a method for generating a fingerprint spectrum according to any of the embodiments of the present disclosure. The non-volatile storage media includes, but is not limited to, disk storage, optical storage, flash memory and the like.

[0101] The processor 620 can be implemented with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, and the like discrete hardware components. Accordingly, each module such as a determining module and a judging module can be implemented by a central processing unit (CPU) running instructions stored in a memory for performing corresponding steps, or by a dedicated circuit for performing corresponding steps.

[0102] The bus 600 can use any of a variety of bus structures. For example, the bus structure includes, but is not limited to, an industry standard architecture (ISA) bus, a micro channel architecture (MCA) bus, a peripheral component interconnect (PCI) bus.

[0103] The computer system 60 can further include an input / output interface 630, a network interface 640, a storage interface 650, and the like. These interfaces 630, 640, 650, and the memory 610 and the processor 620 can be connected through the bus 600. The input / output interface 630 can provide a connection interface for display, mouse, keyboard, and the like input / output devices. The network interface 640 provides a connection interface for various networking devices. The storage interface 650 provides a connection interface for external storage devices such as floppy disks, U disks, SD cards, and the like.

[0104] According to some embodiments of the present disclosure, a fingerprint spectrum generation system is also provided, comprising the fingerprint spectrum generation apparatus according to some embodiments of the present disclosure.

[0105] Figure 7 A user interface of the generation system according to some embodiments of the present disclosure is shown.

[0106] The operation of the user interface will be described below. Figure 7

[0107] First, the secondary spectra of 15 standard reference lipids in the multiple samples are sorted into a matrix format and saved as a csv file. Then, the LipidIN-ReverseLipidomics module of the generation system is entered, the analysis is clicked with the left mouse button, the csv file is uploaded and placed under the data folder, the data.zip is compressed, the upload is selected with the left mouse button, and the upload is waited to be completed. The working is clicked with the left mouse button, the running is waited to be completed, the operation records are clicked with the left mouse button, the result.zip is found, and the running result is downloaded with the left mouse button.

[0108] Here, various aspects of the present disclosure are described with reference to flowcharts and / or block diagrams of the methods, apparatuses and computer program products according to embodiments of the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations thereof, can be implemented by computer readable program instructions.

[0109] These computer readable program instructions can be provided to a processor of a general purpose computer, a special purpose computer, or other programmable device to produce a machine, so that the instructions executed by the processor produce the device that implements the functions specified in one or more blocks of the flowcharts and / or block diagrams.

[0110] These computer readable program instructions can also be stored in a computer readable storage medium, which makes the computer work in a specific way, so as to produce a product, including instructions that implement the functions specified in one or more blocks of the flowcharts and / or block diagrams.

[0111] ​The present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment containing both software and hardware aspects.

[0112] By the fingerprint spectrum generation method, system and device in the above embodiments, and the computer readable storage medium, the generation of the fingerprint spectrum is realized.

[0113] Thus far, the fingerprint spectrum generation method and device according to the present disclosure, and the computer readable storage medium have been described in detail. In order to avoid obscuring the concept of the present disclosure, some details well known in the art are not described. Those skilled in the art can fully understand how to implement the technical solutions disclosed herein according to the above description.

Claims

1. A method for generating a fingerprint spectrum, comprising: determining intra-group features of each spectrum of a plurality of spectra of a target category of lipids, wherein the intra-group features are extracted from each fingerprint spectrum by a first activation layer and a linear layer of a fingerprint spectrum generation model, the fingerprint spectrum generation model simultaneously inputs the plurality of spectra, and realizes feature extraction of the spectra and intra-group feature learning based on a Kolmogrov-Arnold theorem, the fingerprint spectrum generation model learns inherent features of individual spectra in input spectra, maps secondary spectra of the lipids to a fixed low-dimensional space of 512 rows, and cuts data into eight matrices by deflection of two Sigmoid activation functions and a rectified linear unit activation, the data is mapped to a latent spatial dimension by the first activation layer and the linear layer, and feature extraction of the data is realized; determining inter-group features of the plurality of spectra according to the intra-group features of the plurality of spectra, wherein the inter-group features of the plurality of spectra are determined according to the intra-group features of the plurality of spectra by a multi-head attention layer of the fingerprint spectrum generation model; generating a fingerprint spectrum of the target category of lipids according to the intra-group features and the inter-group features; wherein the generating the fingerprint spectrum of the target category of lipids according to the intra-group features and the inter-group features comprises: generating normalized features according to the intra-group features and the inter-group features by a normalization layer of the fingerprint spectrum generation model, generating activation features according to the normalized features by a second activation layer of the fingerprint spectrum generation model, and generating the fingerprint spectrum of the target category of lipids according to the activation features; generating the fingerprint spectrum of the target category of lipids according to the normalized features, wherein the fingerprint spectrum of the target category of lipids is generated according to the activation features by an up-sampling layer of the fingerprint spectrum generation model.

2. The method of generating a fingerprint spectrum according to claim 1, wherein, The plurality of spectra have a first resolution, and the fingerprint spectrum has a second resolution, the second resolution being greater than the first resolution. 3.A device for generating a fingerprint spectrum, comprising: an intra-group feature determination module configured to determine intra-group features of each spectrum of a plurality of spectra of a target category of lipids, wherein the intra-group features are extracted from each fingerprint spectrum by a first activation layer and a linear layer of a fingerprint spectrum generation model, the fingerprint spectrum generation model simultaneously inputs the plurality of spectra, and realizes feature extraction of the spectra and intra-group feature learning based on a Kolmogrov-Arnold theorem, the fingerprint spectrum generation model learns inherent features of individual spectra in input spectra, maps secondary spectra of the lipids to a fixed low-dimensional space of 512 rows, and cuts data into eight matrices by deflection of two Sigmoid activation functions and a rectified linear unit activation, the data is mapped to a latent spatial dimension by the first activation layer and the linear layer, and feature extraction of the data is realized; The inter-group feature determination module is configured to determine the inter-group feature of the plurality of spectral graphs according to the intra-group features of the plurality of spectral graphs, wherein the inter-group feature of the plurality of spectral graphs is determined according to the intra-group features of the plurality of spectral graphs by using a multi-head attention layer of the fingerprint spectral graph generation model; The generation module is configured to generate the fingerprint spectral graph of the target category of lipids according to the intra-group features and the inter-group features, wherein the normalized features are generated according to the intra-group features and the inter-group features by using a normalization layer of the fingerprint spectral graph generation model, the activation features are generated according to the normalized features by using a second activation layer of the fingerprint spectral graph generation model, the fingerprint spectral graph of the target category of lipids is generated according to the activation features, the fingerprint spectral graph of the target category of lipids is generated according to the normalized features, and the fingerprint spectral graph of the target category of lipids is generated according to the activation features by using an up-sampling layer of the fingerprint spectral graph generation model.

4. A fingerprint spectral graph generation device, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute a fingerprint spectral graph generation method according to any one of claims 1 to 2 based on instructions stored in the memory.

5. A fingerprint spectral graph generation system, comprising: a fingerprint spectral graph generation device according to claim 3 or 4.

6. A computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement a fingerprint spectral graph generation method according to any one of claims 1 to 2.

7. A computer program product comprising computer program instructions which, when executed by a processor, implement a fingerprint spectral graph generation method according to any one of claims 1 to 2.

Images