Gene filling method, device, computer equipment and storage medium

Abstract

The present application relates to a gene filling method, device, computer equipment and storage medium. The method obtains a missing gene sequence and inputs the missing gene sequence into a gene filling network for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes at least one gene filling sub-network, and different gene filling sub-networks are used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to a generative adversarial network. The above method implements a gene filling method based on a gene filling network trained by a generative adversarial network, and because the generative adversarial network completes the task based on the idea of ​​mutual game, the gene filling sub-network obtained by training does not need to perform gene filling based on the dynamic linkage relationship between the gene missing position and the gene sample in which it is located, and thus has a low degree of dependence on surrounding genes, which can improve the universality of the gene filling method.

Description

Technical Field

[0001] The present application relates to the field of gene prediction technology, and in particular to a gene filling method, device, computer equipment and storage medium. Background Art

[0002] Human gene analysis is an important task in the medical diagnosis of genetic diseases, and is usually performed in clinics and genetics, such as detecting genetic abnormalities. For this diagnosis, geneticists usually observe and compare specific genes in the mother and father, or compare the patient's gene sequence with the gene of the prototype human. However, the sequencing process is very easy to cause gene data loss, which will bring great challenges to the analysis and research of the whole genome.

[0003] At present, researchers have proposed using gene filling to solve the problem of gene loss. In the process of studying the gene filling method, the gene sequence was first inferred based on the traditional hidden Markov model (HMM) and Monte Carlo (MCMC) machine learning methods to achieve gene filling. Later, with the development of deep learning in recent years, many fully convolutional neural network models have been applied to the gene filling method, such as the U-net model.

[0004] However, all of the above methods have the problem of low gene filling accuracy. Summary of the invention

[0005] Based on this, it is necessary to provide a gene filling method, device, computer equipment and storage medium that can effectively improve the accuracy of gene filling in response to the above technical problems.

[0006] In a first aspect, a gene filling method is provided, the method comprising:

[0007] Obtaining missing gene sequences;

[0008] The missing gene sequence is input into a gene filling network for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes at least one gene filling sub-network, and different gene filling sub-networks are used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to a generative adversarial network.

[0009] The gene filling method provided in the above embodiment obtains the missing gene sequence and inputs the missing gene sequence into the gene filling network for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes at least one gene filling sub-network, and different gene filling sub-networks are used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to a generative adversarial network. The above method not only realizes the gene filling of missing gene sequences with any missing ratio, but also realizes the gene filling method of the gene filling network obtained by training the gene filling network based on the generative adversarial network. Since the generative adversarial network completes the task based on the idea of ​​mutual game, the generative adversarial network can learn any random noise input to obtain a complete gene filling output, so that the gene filling sub-network obtained by training the generative adversarial network does not need to perform gene filling based on the dynamic linkage relationship between the gene missing position and the gene sample in which it is located, compared with the gene filling network obtained by training the convolutional neural network, and thus has a low degree of dependence on surrounding genes, and can be directly applied to gene filling data with complex distribution types, which not only improves the accuracy of gene filling, but also improves the universality of the gene filling method.

[0010] In one embodiment, the gene filling network also includes: an identification sub-network, which is used to identify the missing ratio of the missing gene sequence, determine the missing ratio of the missing gene sequence, and determine the corresponding gene filling sub-network according to the missing ratio, and input the missing gene sequence into the determined gene filling sub-network.

[0011] The gene filling network described in the above embodiment realizes automatic identification of the missing ratio of the input missing gene sequence through the recognition sub-network, and automatically selects the corresponding gene filling sub-network to perform gene filling on the missing gene sequence according to the determined missing ratio. Based on this, the gene filling method provided by the present application is not limited to filling a missing gene sequence with a certain missing ratio, and can be applied to a missing gene sequence with any missing ratio.

[0012] In one embodiment, the process of obtaining the gene filling sub-network in advance according to the generative adversarial network training includes:

[0013] Obtain complete gene sequence samples and missing gene sequence samples with a preset missing ratio;

[0014] Inputting the missing gene sequence sample into the generative subnetwork in the generative adversarial network to obtain a pseudogene sequence sample;

[0015] Inputting the complete gene sequence sample and the pseudogene sequence sample into the discriminant subnetwork in the generative adversarial network to obtain a discriminant result;

[0016] The generative sub-network is trained according to the discrimination result, and the trained generative sub-network is used as the gene filling sub-network.

[0017] The above embodiment is based on the gene filling sub-network trained by the generative adversarial network. Since the training sample data required by the generative adversarial network is small, the limitation of the amount of data is broken. Therefore, the gene filling sub-network for accurately implementing gene filling can be trained based on the gene sequence of a small sample, which reduces the difficulty of implementing gene filling to a certain extent. Moreover, the generative adversarial network does not have high requirements on the form of input data, and is more universal. Therefore, the gene filling sub-network trained by the generative adversarial network is still robust to the filling operation of various types of gene sequences.

[0018] In one embodiment, the step of obtaining a complete gene sequence sample and a missing gene sequence sample with a preset missing ratio includes:

[0019] Obtain gene sequences of different species from a preset gene database as a sample set;

[0020] The complete gene sequence sample and the missing gene sequence sample with a preset missing ratio are obtained according to the sample set.

[0021] The above embodiment can directly obtain missing gene sequence samples through the gene database, and use the existing missing gene sequence samples for training, which can improve the training effect to a certain extent.

[0022] In one embodiment, obtaining the complete gene sequence sample and the missing gene sequence sample with a preset missing ratio according to the sample set includes:

[0023] Taking the gene sequence in the sample set as the complete gene sequence sample;

[0024] According to the preset missing ratio, random missing processing is performed on the gene sequences in the sample set to obtain the missing gene sequence samples with the preset missing ratio.

[0025] The missing gene sequence samples in this embodiment are generated from complete gene sequence samples, so the missing gene sequence samples and the complete gene sequence samples are completely corresponding, which can improve the training effect of the gene filling network based on the missing gene sequence samples and the complete gene sequence samples in the later stage, so that the gene filling sub-network can perform gene filling more accurately.

[0026] In one embodiment, obtaining the complete gene sequence sample and the missing gene sequence sample with a preset missing ratio according to the sample set includes:

[0027] According to the preset missing ratio, a preset number of gene sequences in the sample set are randomly deleted to obtain missing gene sequence samples with a preset missing ratio;

[0028] The remaining gene sequences in the sample set are taken as the complete gene sequence samples.

[0029] The above embodiment obtains missing gene sequence samples by randomly missing complete gene sequences in the gene database. This method solves the problem that it is difficult to obtain missing gene sequence samples due to the small number of missing gene sequences in the existing gene database, and improves the feasibility and practicality of the above method of training a generative adversarial network based on missing gene sequence samples.

[0030] In one embodiment, after obtaining the missing gene sequence, the method further comprises:

[0031] Encoding the missing gene sequence according to a preset encoding method to generate matrix data corresponding to the missing gene sequence;

[0032] The step of inputting the missing gene sequence into a gene filling network for gene filling to obtain a complete gene sequence comprises:

[0033] The matrix data is input into a gene filling network for gene filling to obtain a complete gene sequence.

[0034] The above embodiment encodes the missing gene sequence so that the missing gene sequence can be converted into format data that can be recognized by the gene filling network, so as to improve the accuracy of the gene filling network in performing gene filling on the missing gene sequence.

[0035] In one embodiment, the step of inputting the matrix data into a gene filling network for gene filling to obtain a complete gene sequence comprises:

[0036] Inputting the matrix data into a gene filling network for gene filling to obtain matrix data corresponding to a complete gene sequence;

[0037] The matrix data corresponding to the complete gene sequence is decoded according to a preset decoding method to generate the complete gene sequence; the preset decoding method corresponds to the preset encoding method.

[0038] The above embodiment decodes the data output by the gene filling network so that the data output by the gene filling network can be converted into a gene sequence, so as to obtain a complete gene sequence corresponding to the missing gene sequence.

[0039] In a second aspect, a gene filling device is provided, the device comprising:

[0040] An acquisition module, used to obtain missing gene sequences;

[0041] A gene filling module is used to input the missing gene sequence into a gene filling network for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes at least one gene filling sub-network, and different gene filling sub-networks are used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to a generative adversarial network.

[0042] In a third aspect, a computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the method described in the first aspect when executing the computer program.

[0043] In a fourth aspect, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is an internal structure diagram of a computer device in one embodiment;

[0045] Figure 2 A schematic diagram of a gene filling method according to an embodiment;

[0046] Figure 3 A schematic diagram of the structure of a gene filling network in one embodiment;

[0047] Figure 4 A schematic diagram of the structure of a gene filling network in one embodiment;

[0048] Figure 5 A schematic diagram of a gene filling method according to an embodiment;

[0049] Figure 6 A schematic diagram of a gene filling method according to an embodiment;

[0050] Figure 7 A schematic diagram of a process for generating a schematic diagram of a structure of an adversarial network in one embodiment;

[0051] Figure 8 A schematic diagram of a gene filling method according to an embodiment;

[0052] Fig. 9 A schematic diagram of a gene filling method according to an embodiment;

[0053] Fig.10 A schematic diagram of a gene filling method according to an embodiment;

[0054] Fig.11 A schematic diagram of a gene filling method according to an embodiment;

[0055] Fig.12 A schematic diagram of a gene filling method according to an embodiment;

[0056] Fig.13 A schematic diagram of a gene filling method according to an embodiment;

[0057] Fig.14 is a structural block diagram of a gene filling device in one embodiment;

[0058] Fig.15 is a structural block diagram of a gene filling device in one embodiment;

[0059] Fig.16 is a structural block diagram of a gene filling device in one embodiment;

[0060] Fig.17 FIG. 4 is a structural block diagram of a gene filling device in one embodiment. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0062] The gene filling method provided in this application can be applied to Figure 1 In the computer device shown in the figure, the computer device may be a server, or the computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 1 The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities.

[0063] The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by a processor, a gene filling method is implemented.

[0064] The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0065] Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0066] In one embodiment, Figure 2 As shown, a gene filling method is provided, which is applied to Figure 1 The computer device in the example is used to illustrate, including the following steps:

[0067] S101, obtaining the missing gene sequence.

[0068] Among them, the missing gene sequence is the gene sequence to be filled, and specifically can be a gene sequence with a certain missing ratio, and the missing ratio can be any ratio, for example, a 10% missing gene sequence, a 20% missing gene sequence, a 30% missing gene sequence, etc.

[0069] In this embodiment, the computer device can obtain the missing gene sequence from any gene database, or obtain the missing gene sequence from the gene experimental research process, so as to fill the missing gene sequence later to obtain a complete gene sequence and promote experimental research. The above gene database can be a public database of any species, such as the Animal-ImputeDB public database.

[0070] S102, inputting the missing gene sequence into the gene filling network for gene filling to obtain a complete gene sequence.

[0071] Among them, Figure 3 As shown, the gene filling network includes at least one gene filling sub-network, and if it includes multiple gene filling sub-networks, the multiple gene filling sub-networks are arranged in parallel, each gene filling sub-network is used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to the generative adversarial network.

[0072] In this embodiment, the computer device can pre-construct a generative adversarial network, and train the constructed generative adversarial network based on the missing gene sequence samples and complete gene sequence samples with preset missing ratios, and then obtain a gene filling subnetwork according to the trained generative adversarial network, and the gene filling subnetwork obtained by the training can be used to perform gene filling on the missing gene sequence with preset missing ratios, and then obtain the corresponding complete gene sequence. It should be noted that when the gene filling network includes multiple gene filling subnetworks, the computer device can pre-construct multiple generative adversarial networks, and respectively train multiple gene filling subnetworks constructed based on missing gene sequence samples and complete gene sequence samples with different preset missing ratios, then multiple gene filling subnetworks can be obtained according to the trained multiple generative adversarial networks, and multiple gene filling subnetworks can be used to fill missing gene sequences with different missing ratios, and then output the complete gene sequence corresponding to the missing gene sequences with different missing ratios. When the computer device trains the generative adversarial network to obtain multiple gene filling subnetworks, the multiple gene filling subnetworks can be used to form a gene filling network together.

[0073] In practical applications, when a computer device obtains a missing gene sequence, if the missing gene sequence is a missing gene sequence with a certain missing ratio, the missing gene sequence is input into the gene filling network, so that the gene filling sub-network for the corresponding missing ratio in the gene filling network performs gene filling on the missing gene sequence to obtain a complete gene sequence corresponding to the missing gene sequence. For example, if the missing gene sequence is a missing gene sequence with a missing ratio of 10%, then the following is used: Figure 3 The 10% gene filling subnetwork described in performs gene filling on the missing gene sequence. Optionally, when the missing gene sequence acquired by the computer device is a plurality of missing gene sequences with different missing ratios, the computer device can simultaneously input the plurality of missing gene sequences into the gene filling network, and each gene filling subnetwork in the gene filling network performs gene filling on the missing gene sequences with different missing ratios respectively, to obtain the complete gene sequence corresponding to each missing gene sequence.

[0074] The gene filling method provided in the above embodiment obtains the missing gene sequence and inputs the missing gene sequence into the gene filling network for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes at least one gene filling sub-network, and different gene filling sub-networks are used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to a generative adversarial network. The above method not only realizes the gene filling of missing gene sequences with any missing ratio, but also realizes the gene filling method of the gene filling network obtained by training the gene filling network based on the generative adversarial network, and because the generative adversarial network completes the task based on the idea of ​​mutual game, the generative adversarial network can learn any random noise input to obtain a complete gene filling output, so that the gene filling sub-network obtained by training the generative adversarial network does not need to perform gene filling based on the dynamic linkage relationship between the gene missing position and the gene sample in which it is located, compared with the gene filling network obtained by training the convolutional neural network, and thus has a low degree of dependence on surrounding genes, and can be directly applied to gene filling data with complex distribution types, which not only improves the accuracy of gene filling, but also improves the universality of the gene filling method.

[0075] Based on the above Figure 3 The gene filling network structure described above, the present application also provides a gene filling network, such as Figure 4 As shown, the gene filling network also includes: an identification subnetwork, which is used to identify the missing gene sequence, determine the missing ratio of the missing gene sequence, and determine the corresponding gene filling subnetwork according to the missing ratio, and input the missing gene sequence into the determined gene filling subnetwork.

[0076] In this embodiment, when the computer device obtains a missing gene sequence, it does not know the missing ratio of the missing gene sequence. Therefore, the missing gene sequence can be input into the identification sub-network in the gene filling network to identify the missing ratio, obtain the missing ratio of the missing gene sequence, and then determine the corresponding gene filling sub-network according to the missing ratio, and input the missing gene sequence into the determined gene filling sub-network for gene filling. It should be noted that when the computer device obtains multiple missing gene sequences, and the multiple missing gene sequences are missing gene sequences with different missing ratios, the computer device can sequentially input the multiple missing gene sequences into the identification sub-network, so that the identification sub-network sequentially identifies the missing ratios of different missing gene sequences, and after selecting different gene filling sub-networks according to different missing ratios, the different missing gene sequences are input into their respective corresponding gene filling sub-networks for gene filling, and obtain the complete gene sequence corresponding to each missing gene sequence.

[0077] The gene filling network described in the above embodiment realizes automatic identification of the missing ratio of the input missing gene sequence through the recognition sub-network, and automatically selects the corresponding gene filling sub-network to perform gene filling on the missing gene sequence according to the determined missing ratio. Based on this, the gene filling method provided by the present application is not limited to filling a missing gene sequence with a certain missing ratio, and can be applied to a missing gene sequence with any missing ratio.

[0078] In practical applications, when a computer device obtains a missing gene sequence, it can also pre-process the missing gene sequence so that the missing gene sequence can be converted into format data that can be recognized by the gene filling network, such as vector data or matrix data. Therefore, a method for pre-processing the gene sequence is also provided. Figure 5 As shown, Figure 2 The method described in the embodiment also includes:

[0079] S103, encoding the missing gene sequence according to a preset encoding method to generate matrix data corresponding to the missing gene sequence.

[0080] The preset encoding method is used to convert the gene sequence into matrix data or vector data. The preset encoding method can be any encoding method that can convert the gene sequence and is not limited here.

[0081] This embodiment involves a process of preprocessing the missing gene sequence, that is, when the computer device obtains the original missing gene sequence, it is necessary to encode the missing gene sequence according to a preset encoding method to generate matrix data corresponding to the missing gene sequence. For example, the missing gene sequence is arranged into four different data combinations according to four types of four-dimensional binary data encoding methods, and presented as matrix data.

[0082] Correspondingly, the above step S102 includes: inputting the matrix data into the gene filling network for gene filling to obtain a complete gene sequence.

[0083] The specific method of using the gene filling network to perform gene filling on the input data is described in the above S102. Please refer to the above description for details, which will not be repeated here.

[0084] The above embodiment encodes the missing gene sequence so that the missing gene sequence can be converted into format data that can be recognized by the gene filling network, so as to improve the accuracy of the gene filling network in performing gene filling on the missing gene sequence.

[0085] Correspondingly, the computer device generates matrix data after encoding the missing gene sequence, and then inputs the matrix data into the gene filling network for processing. Then, the data output by the gene filling network is naturally matrix data. In this case, it is also necessary to decode the matrix data output by the gene filling network to obtain the complete gene sequence in the original gene format. Therefore, the above step S102, such as Figure 6 As shown, including:

[0086] S1021, inputting the matrix data into the gene filling network for gene filling to obtain the matrix data corresponding to the complete gene sequence.

[0087] This step is basically the same as the above step S102. For details, please refer to the above description and will not be repeated here.

[0088] S1022, decoding the matrix data corresponding to the complete gene sequence according to a preset decoding method to generate a complete gene sequence.

[0089] Among them, the preset decoding method is used to convert matrix data or vector data into a gene sequence, and the preset decoding method corresponds to the previous preset encoding method.

[0090] This embodiment involves a process of post-processing the data output by the gene filling network, that is, when a computer device obtains matrix data corresponding to a complete gene sequence output by the gene filling network, it is necessary to decode the matrix data output by the gene filling network according to a preset decoding method to generate a complete gene sequence corresponding to the matrix data.

[0091] The above embodiment decodes the data output by the gene filling network so that the data output by the gene filling network can be converted into a gene sequence, so as to obtain a complete gene sequence corresponding to the missing gene sequence.

[0092] In one embodiment, the present application provides a method for training each gene filling subnetwork as described above, which can be applied to Figure 7 The generative adversarial network shown in Figure 8 As shown, the training method includes:

[0093] S201, obtaining a complete gene sequence sample and a missing gene sequence sample with a preset missing ratio.

[0094] Among them, the preset missing ratio can be determined in advance by the computer equipment according to the actual training needs. For example, if the gene filling sub-network obtained through training needs to have the best gene filling effect for a gene sequence with 10% missing, the preset missing ratio is set to 10%; if the gene filling sub-network obtained through training needs to have the best gene filling effect for a gene sequence with 20% missing, the preset missing ratio is set to 20%; if the gene filling sub-network obtained through training needs to have the best gene filling effect for a gene sequence with 30% missing, the preset missing ratio is set to 30%.

[0095] In this embodiment, gene sequences of different species are obtained from any public database, such as chicken, duck, pig and dog gene sequences, to generate a large number of gene sequence samples. Generally, the gene sequences obtained from the public database are complete gene sequences, but due to improper storage or usage factors, it is not ruled out that there are missing gene sequences. Therefore, when the computer device obtains a large number of gene sequence samples from the public database, some complete gene sequences can be screened out as complete gene sequence samples, and the corresponding number of missing gene sequences with a preset missing ratio can be used as missing gene sequence samples. Since the generative adversarial network does not require a high amount of data, even if the sample amount of complete gene sequence samples and missing gene sequence samples that can be obtained based on the above method is small, it will not affect the later use of small sample amount sample data to train the generative adversarial network to obtain a gene filling network.

[0096] S202, inputting the missing gene sequence sample into the generative sub-network in the generative adversarial network to obtain a pseudo gene sequence sample.

[0097] Among them, the pseudo gene sequence sample is a fake complete gene sequence. The generative adversarial network consists of a generative subnetwork and a discriminative subnetwork (see Figure 7 The generative adversarial network shown in FIG. 1 is used to fill in the missing gene sequence samples with a preset missing ratio of the input genes, and generate pseudogene sequence samples. The discriminative subnetwork is used to determine whether the pseudogene sequence samples are close to the real complete gene sequence.

[0098] In this embodiment, when the computer device obtains a missing gene sequence sample with a preset missing ratio, the missing gene sequence sample can be preprocessed first. Specifically, the missing gene sequence sample can be genetically encoded in a preset encoding method, and the missing gene sequence in the missing gene sequence sample can be converted into matrix data corresponding to the gene sequence sample according to a certain arrangement format, and then the matrix data corresponding to the gene sequence sample is input into the generation sub-network to obtain a pseudo gene sequence sample. For example, the missing gene sequence is genetically encoded to generate 16*16 matrix data as the input of the generation sub-network, and the channel of the generation sub-network can be set to 128. Specifically, all the missing gene sequences contained in the missing gene sequence sample can be preprocessed at one time and then input into the generation sub-network, or all the missing gene sequences contained in the missing gene sequence sample can be preprocessed successively and then input into the generation sub-network in sequence, and of course, one missing gene sequence contained in the missing gene sequence sample can be preprocessed at one time and then input into the generation sub-network.

[0099] S203, inputting the complete gene sequence sample and the pseudo gene sequence sample into the discriminant sub-network in the generative adversarial network to obtain a discrimination result.

[0100] When the generating subnetwork outputs a pseudogene sequence sample, the computer device can further input the pseudogene sequence sample and the previously acquired complete gene sequence sample into the discriminating subnetwork for discrimination, specifically discriminating whether the pseudogene sequence sample is close to the complete gene sequence sample, and obtaining a discrimination result. The discrimination result can indicate the degree to which the pseudogene sequence sample is close to the complete gene sequence sample, which can be represented by a numerical value of [0-1], where 1 represents that the pseudogene sequence sample is completely close to the complete gene sequence sample, that is, it reaches the degree of being indistinguishable from the real thing; 0 represents that the pseudogene sequence sample is completely different from the complete gene sequence sample, that is, the pseudogene sequence sample is not a complete gene sequence. It should be noted that before the computer device simultaneously inputs the pseudogene sequence sample and the previously acquired complete gene sequence sample into the discriminating subnetwork, it is also necessary to encode the complete gene sequence sample and the pseudogene sequence sample in a preset encoding method, specifically, converting the complete gene sequence sample and the pseudogene sequence sample into matrix data and then inputting them into the discriminating subnetwork.

[0101] S204, training a generative sub-network according to the discrimination result, and using the trained generative sub-network as a gene filling sub-network.

[0102] When the computer device obtains the discrimination result, the target loss function is set, and the discrimination result is substituted into the target loss function to calculate the loss value, and then the generative subnetwork and the discriminative subnetwork are alternately trained according to the loss value to obtain the trained generative subnetwork and the discriminative subnetwork, and then the trained generative subnetwork is used as the gene filling subnetwork. It should be noted that the above-mentioned alternating training process can be performed by training the discriminative subnetwork once, and the generative subnetwork is trained multiple times based on the once trained discriminative subnetwork, and the training is repeated until the loss value reaches the preset requirement, that is, the training of the generative subnetwork is completed; optionally, the above-mentioned alternating training process can also train the generative subnetwork and the discriminative subnetwork at the same time with equal frequency, and the training is repeated until the loss value reaches the preset requirement, that is, the training of the generative subnetwork is completed.

[0103] It is understandable that the above Figure 8 The embodiment is a process of training a gene filling sub-network. Figure 2 When the gene filling network proposed in the embodiment includes multiple gene filling sub-networks, the following can be used: Figure 8 The method described in the embodiment is based on multiple generative adversarial network training to obtain multiple gene filling sub-networks. The training process of each gene filling sub-network is basically the same and can be trained independently. It only needs to set different preset missing gene sequence samples with different missing ratios to train and obtain gene filling sub-networks for filling missing gene sequences with different missing ratios.

[0104] The above embodiment is based on the gene filling sub-network trained by the generative adversarial network. Since the training sample data required by the generative adversarial network is small, the limitation of the amount of data is broken. Therefore, the gene filling sub-network for accurately implementing gene filling can be trained based on the gene sequence of a small sample, which reduces the difficulty of implementing gene filling to a certain extent. Moreover, the generative adversarial network does not have high requirements on the form of input data, and is more universal. Therefore, the gene filling sub-network trained by the generative adversarial network is still robust to the filling operation of various types of gene sequences.

[0105] Furthermore, an implementation of the above S201 is provided, such as Fig. 9 As shown, the above S201 "obtaining a complete gene sequence sample and a missing gene sequence sample with a preset missing ratio" includes:

[0106] S301, obtaining gene sequences of different species from a preset gene database as a sample set.

[0107] The preset gene database in this embodiment can be an Animal-ImputeDB public database, which stores complete gene sequences and missing gene sequences with different missing ratios, and stores gene sequences of different species in the preset gene database. The computer device can select a certain number of gene sequences from the preset gene database and download them as a sample set for use in training. The selected gene sequences can be all complete gene sequences, or they can include both some complete gene sequences and some missing gene sequences.

[0108] S302, obtaining complete gene sequence samples and missing gene sequence samples with a preset missing ratio according to the sample set.

[0109] When the computer device downloads a sample set from a preset gene database, and the sample set contains both complete gene sequences and missing gene sequences, the computer device can directly use the complete gene sequence from the sample set as a complete gene sequence sample, screen the missing gene sequence again, screen out the missing gene sequence with a preset missing ratio, and directly use the missing gene sequence with the preset missing ratio as the missing gene sequence sample. This embodiment can directly obtain the missing gene sequence sample through the gene database, and use the existing missing gene sequence samples for training, which can improve the training effect to a certain extent.

[0110] Optionally, when the sample set contains complete gene sequences, two implementation methods are provided to obtain complete gene sequence samples and missing gene sequence samples with a preset missing ratio:

[0111] That is, the first implementation method, such as Fig.10 As shown, this is achieved through the following steps:

[0112] S401, randomly performing deletion processing on a preset number of gene sequences in a sample set according to a preset deletion ratio to obtain missing gene sequence samples with a preset deletion ratio.

[0113] Among them, the preset missing ratio can be determined in advance by the computer device according to the actual training needs. For example, if the gene filling network obtained by training needs to have the best gene filling effect for a gene sequence with 10% missing, the preset missing ratio is set to 10%; if the gene filling network obtained by training needs to have the best gene filling effect for a gene sequence with 20% missing, the preset missing ratio is set to 20%; if the gene filling network obtained by training needs to have the best gene filling effect for a gene sequence with 30% missing, the preset missing ratio is set to 30%. The preset number can be determined in advance by the computer device according to the actual training needs and the sample set size, which is not limited here.

[0114] In this embodiment, when the computer device obtains the sample set, some gene sequences can be screened out therefrom, specifically, half of the gene sequences can be screened out, random deletion processing is performed according to a preset deletion ratio, and a missing gene sequence sample with a preset deletion ratio is generated according to the screened gene sequence. For example, a missing gene sequence sample with 10% deletion is obtained after the screened gene sequence is randomly deleted.

[0115] S402, taking the remaining gene sequences in the sample set as complete gene sequence samples.

[0116] In this embodiment, when the computer device obtains missing gene sequence samples according to the preset number of gene sequences in the sample set according to the above steps, the remaining gene sequences in the sample set can be used as complete gene sequence samples, for example, the other half of the gene sequences in the sample set can be used as complete gene sequence samples. Because the sample set in this embodiment contains complete gene sequences, the remaining gene sequences can be used as complete gene sequence samples.

[0117] The above embodiment obtains missing gene sequence samples by randomly missing complete gene sequences in the gene database. This method solves the problem that it is difficult to obtain missing gene sequence samples due to the small number of missing gene sequences in the existing gene database, and improves the feasibility and practicality of the above method of training a generative adversarial network based on missing gene sequence samples.

[0118] The second implementation method is Fig.11 As shown, this is achieved through the following steps:

[0119] S501, taking the gene sequence in the sample set as a complete gene sequence sample.

[0120] Because the sample set in this embodiment contains complete gene sequences, the gene sequences in the sample set can be directly used as complete gene sequence samples.

[0121] S502, performing random deletion processing on the gene sequences in the sample set according to a preset deletion ratio to obtain a missing gene sequence sample with a preset deletion ratio.

[0122] In this embodiment, when the computer device obtains a complete gene sequence sample, the complete gene sequence sample is randomly deleted according to a preset deletion ratio to obtain a deleted gene sequence sample with a preset deletion ratio. The deleted gene sequence sample in this embodiment is generated from a complete gene sequence sample, so the deleted gene sequence sample and the complete gene sequence sample are completely corresponding, which can improve the training effect of the gene filling network based on the deleted gene sequence sample and the complete gene sequence sample in the later stage, so that the gene filling sub-network can perform gene filling more accurately.

[0123] Based on all the above embodiments, the present application also provides a gene filling method, which is applied to Figure 4 The gene filling network shown is Fig.12 As shown, the method includes:

[0124] S601, obtaining the missing gene sequence.

[0125] S602, encoding the missing gene sequence according to a preset encoding method to generate matrix data corresponding to the missing gene sequence.

[0126] S603, inputting the matrix data corresponding to the missing gene sequence into the recognition sub-network to identify the missing ratio, and obtaining the missing ratio of the missing gene sequence.

[0127] S604, determining a corresponding gene filling sub-network according to the missing ratio of the missing gene sequence.

[0128] S605, inputting the matrix data corresponding to the missing gene sequence into the determined gene filling subnetwork for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence.

[0129] The above steps have been explained above. Please refer to the above instructions for details and will not be repeated here.

[0130] Optionally, the present application also provides a gene filling method, which is applied to Figure 4 The gene filling network shown is Fig.13 As shown, the method includes:

[0131] S701, obtaining multiple deleted gene sequences with different deletion ratios.

[0132] S702, encoding the multiple missing gene sequences respectively according to a preset encoding method to generate matrix data corresponding to the multiple missing gene sequences.

[0133] S703, inputting the matrix data corresponding to the multiple missing gene sequences into the recognition sub-network to respectively identify the missing ratios, and obtaining the missing ratios of the respective missing gene sequences.

[0134] S704, determining the corresponding gene filling sub-network according to the missing ratio of each missing gene sequence.

[0135] S705, inputting the matrix data corresponding to the multiple missing gene sequences into the determined corresponding gene filling sub-networks respectively for gene filling, and obtaining the complete gene sequence corresponding to each missing gene sequence.

[0136] The above steps have been explained above. Please refer to the above instructions for details and will not be repeated here.

[0137] Fig.12 The embodiment realizes the gene filling of a missing gene sequence with a certain missing ratio. Fig.13 The embodiment realizes the gene filling of multiple missing gene sequences with different missing ratios. Therefore, the gene filling method proposed in the present application can be applicable to the filling needs of missing gene sequences with multiple missing ratios, and has extremely high universality. Moreover, since each gene filling sub-network contained in the gene filling network is obtained by training with different generative adversarial networks according to training samples with different preset missing ratios, each gene filling sub-network has the best filling effect on the missing gene sequences with their own missing ratios, so the gene filling network proposed in the present application has extremely high accuracy in filling missing gene sequences with various missing ratios.

[0138] It should be understood that although Figure 2-13 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2-13 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0139] In one embodiment, Fig.14 As shown, a gene filling device is provided, comprising:

[0140] The acquisition module 11 is used to acquire the missing gene sequence.

[0141] The gene filling module 12 is used to input the missing gene sequence into the gene filling network for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes at least one gene filling sub-network, and different gene filling sub-networks are used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to a generative adversarial network.

[0142] In one embodiment, the gene filling network also includes: an identification subnetwork, which is used to identify the missing ratio of the missing gene sequence, determine the missing ratio of the missing gene sequence, and determine the corresponding gene filling subnetwork according to the missing ratio, and input the missing gene sequence into the determined gene filling subnetwork.

[0143] In one embodiment, Fig.15 As shown, the above device also includes: a training module 13, the training module 13 includes:

[0144] A first acquisition unit 131 is used to acquire a complete gene sequence sample and a missing gene sequence sample with a preset missing ratio;

[0145] A second acquisition unit 132 is used to input the missing gene sequence sample into the generation sub-network in the generative adversarial network to obtain a pseudo gene sequence sample;

[0146] A third acquisition unit 133 is used to input the complete gene sequence sample and the pseudo gene sequence sample into the discriminant subnetwork in the generative adversarial network to obtain a discrimination result;

[0147] The training unit 134 is used to train the generation sub-network according to the discrimination result, and use the trained generation sub-network as the gene filling sub-network.

[0148] In one embodiment, Fig.16 As shown, the first acquisition unit 131 includes:

[0149] An acquisition subunit 1311 is used to acquire gene sequences of different species from a preset gene database as a sample set;

[0150] The determination subunit 1312 is used to obtain the complete gene sequence sample and the missing gene sequence sample with a preset missing ratio according to the sample set.

[0151] In one embodiment, the acquisition subunit 1311 is specifically used to use the gene sequence in the sample set as the complete gene sequence sample; and perform random deletion processing on the gene sequence in the sample set according to the preset deletion ratio to obtain the missing gene sequence sample with the preset deletion ratio.

[0152] In one embodiment, the acquisition subunit 1311 is specifically used to perform random deletion processing on a preset number of gene sequences in the sample set according to the preset deletion ratio to obtain missing gene sequence samples with a preset deletion ratio; and use the remaining gene sequences in the sample set as the complete gene sequence samples.

[0153] In one embodiment, Fig.17 As shown, the above device also includes:

[0154] An encoding module 13, used to encode the missing gene sequence according to a preset encoding method to generate matrix data corresponding to the missing gene sequence;

[0155] Correspondingly, the above-mentioned gene filling module 12 is specifically used to input the matrix data into the gene filling network for gene filling to obtain a complete gene sequence; it is also specifically used to input the matrix data into the gene filling network for gene filling to obtain the matrix data corresponding to the complete gene sequence; according to a preset decoding method, the matrix data corresponding to the complete gene sequence is decoded to generate the complete gene sequence; the preset decoding method corresponds to the preset encoding method.

[0156] For the specific definition of the gene filling device, please refer to the definition of the gene filling method above, which will not be repeated here. Each module in the above-mentioned gene filling device can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0157] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0158] Obtaining missing gene sequences;

[0159] The missing gene sequence is input into a gene filling network for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes at least one gene filling sub-network, and different gene filling sub-networks are used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to a generative adversarial network.

[0160] The above embodiment provides a computer device, whose implementation principle and technical effect are similar to those of the above method embodiment, and will not be repeated here.

[0161] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0162] Obtaining missing gene sequences;

[0163] The missing gene sequence is input into a gene filling network for gene filling to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes at least one gene filling sub-network, and different gene filling sub-networks are used to fill missing gene sequences with different missing ratios, and each gene filling sub-network is pre-trained according to a generative adversarial network.

[0164] The above embodiment provides a computer-readable storage medium, whose implementation principle and technical effect are similar to those of the above method embodiment, and will not be repeated here.

[0165] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0166] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0167] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A gene filling method, characterized in that: The method comprises: Obtaining a missing gene sequence; the missing gene sequence is a gene sequence with a missing ratio; According to the missing ratio of the missing gene sequence, the missing gene sequence is input into a gene filling subnetwork of the corresponding missing ratio in the gene filling network for gene filling, so as to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes a plurality of gene filling subnetworks, and different gene filling subnetworks are used to fill missing gene sequences of different missing ratios, and each gene filling subnetwork is pre-trained according to a generative adversarial network; The process of obtaining the gene filling sub-network in advance according to the generative adversarial network training includes: Obtain complete gene sequence samples and missing gene sequence samples with a preset missing ratio; Inputting the missing gene sequence sample into the generative subnetwork in the generative adversarial network to obtain a pseudogene sequence sample; Inputting the complete gene sequence sample and the pseudogene sequence sample into the discriminant subnetwork in the generative adversarial network to obtain a discriminant result; The generative sub-network is trained according to the discrimination result, and the trained generative sub-network is used as the gene filling sub-network.

2. The method according to claim 1, characterized in that The gene filling network also includes: an identification subnetwork, which is used to identify the missing ratio of the missing gene sequence, determine the missing ratio of the missing gene sequence, and determine the corresponding gene filling subnetwork according to the missing ratio, and input the missing gene sequence into the determined gene filling subnetwork.

3. The method according to claim 2, characterized in that The method of obtaining a complete gene sequence sample and a missing gene sequence sample with a preset missing ratio comprises: Obtain gene sequences of different species from a preset gene database as a sample set; The complete gene sequence sample and the missing gene sequence sample with a preset missing ratio are obtained according to the sample set.

4. The method according to claim 3, characterized in that The step of obtaining the complete gene sequence sample and the missing gene sequence sample with a preset missing ratio according to the sample set includes: Taking the gene sequence in the sample set as the complete gene sequence sample; According to the preset missing ratio, random missing processing is performed on the gene sequences in the sample set to obtain the missing gene sequence samples with the preset missing ratio.

5. The method according to claim 3, characterized in that: The step of obtaining the complete gene sequence sample and the missing gene sequence sample with a preset missing ratio according to the sample set includes: According to the preset missing ratio, a preset number of gene sequences in the sample set are randomly deleted to obtain missing gene sequence samples with a preset missing ratio; The remaining gene sequences in the sample set are taken as the complete gene sequence samples.

6. The method according to claim 1 or 2, characterized in that: After obtaining the missing gene sequence, the method further comprises: Encoding the missing gene sequence according to a preset encoding method to generate matrix data corresponding to the missing gene sequence; The step of inputting the missing gene sequence into a gene filling network for gene filling to obtain a complete gene sequence comprises: The matrix data is input into a gene filling network for gene filling to obtain a complete gene sequence.

7. The method according to claim 6, characterized in that The step of inputting the matrix data into a gene filling network for gene filling to obtain a complete gene sequence comprises: Inputting the matrix data into a gene filling network for gene filling to obtain matrix data corresponding to a complete gene sequence; The matrix data corresponding to the complete gene sequence is decoded according to a preset decoding method to generate the complete gene sequence; the preset decoding method corresponds to the preset encoding method.

8. A gene filling device, characterized in that: The device comprises: An acquisition module, used to acquire a missing gene sequence; the missing gene sequence is a gene sequence with a missing ratio; A gene filling module, used for inputting the missing gene sequence into a gene filling subnetwork with a corresponding missing ratio in a gene filling network to perform gene filling according to the missing ratio of the missing gene sequence, so as to obtain a complete gene sequence corresponding to the missing gene sequence, wherein the gene filling network includes a plurality of gene filling subnetworks, and different gene filling subnetworks are used to fill missing gene sequences with different missing ratios, and each gene filling subnetwork is pre-trained according to a generative adversarial network; The process of obtaining the gene filling sub-network in advance according to the generative adversarial network training includes: Obtain complete gene sequence samples and missing gene sequence samples with a preset missing ratio; Inputting the missing gene sequence sample into the generative subnetwork in the generative adversarial network to obtain a pseudogene sequence sample; Inputting the complete gene sequence sample and the pseudogene sequence sample into the discriminant subnetwork in the generative adversarial network to obtain a discriminant result; The generative sub-network is trained according to the discrimination result, and the trained generative sub-network is used as the gene filling sub-network.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images