Methods, devices, equipment, media and products for processing transcription sequencing data

Abstract

This invention discloses a method, apparatus, device, medium, and product for processing transcription sequencing data. The method includes: acquiring transcription sequencing training sets from at least two sequencing platforms; inputting the transcription sequencing training sets into a generator in a generative adversarial network (GAN) to obtain an output reconstructed sequencing dataset; determining a first platform label set corresponding to the transcription sequencing training set and a second platform label set corresponding to the reconstructed sequencing dataset using a discriminator in the GAN; and iteratively training the generator and discriminator alternately based on the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set to obtain a trained GAN. The generator in the trained GAN is used to reconstruct the transcription sequencing data to be tested from any sequencing platform. This invention improves the elimination of sequencing bias across platforms.

Description

Technical Field

[0001] This invention relates to the fields of genomics and bioinformatics, and in particular to a method, apparatus, equipment, medium, and product for processing transcription sequencing data. Background Technology

[0002] In oncology research, high-throughput sequencing technology has become a key means to reveal the molecular characteristics of tumors. By analyzing the gene expression profile of tumor tissues, we can gain a deeper understanding of the molecular mechanisms of tumor occurrence and development, and it also helps to identify diagnostic biomarkers and therapeutic targets.

[0003] However, due to significant differences in technical principles, probe design, experimental procedures, and data analysis among different sequencing platforms, the transcriptional sequencing data obtained from different sequencing platforms suffer from severe batch effects and distribution inconsistencies.

[0004] Although existing methods attempt to eliminate sequencing bias across platforms, these methods either rely on prior knowledge and are difficult to handle unknown confounding factors, or fail to effectively capture complex nonlinear platform differences due to linear assumptions. Therefore, the de-platform differentiation effect of traditional methods is not ideal. Summary of the Invention

[0005] This invention provides a method, apparatus, device, medium, and product for processing transcription sequencing data to address the problem of poor platform differentiation removal in traditional methods. It eliminates cross-platform sequencing bias while preserving the biological characteristic information contained in the transcription sequencing data.

[0006] According to an embodiment of the present invention, a method for processing transcription sequencing data is provided, the method comprising:

[0007] Obtain transcription sequencing training sets from at least two sequencing platforms;

[0008] The transcriptional sequencing training set is input into the generator in the generative adversarial network to obtain the output reconstructed sequencing dataset;

[0009] The discriminator in the generative adversarial network determines a first platform label set corresponding to the transcription sequencing training set and a second platform label set corresponding to the reconstructed sequencing dataset.

[0010] Based on the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set, the generator and the discriminator are alternately iteratively trained to obtain a trained generative adversarial network;

[0011] The generator in the trained generative adversarial network is used to reconstruct the transcriptional sequencing data to be tested from any sequencing platform.

[0012] According to another embodiment of the present invention, a transcription sequencing data processing apparatus is provided, the apparatus comprising:

[0013] The transcription sequencing training set acquisition module is used to acquire transcription sequencing training sets from at least two sequencing platforms;

[0014] The reconstructed sequencing dataset output module is used to input the transcription sequencing training set into the generator in the generative adversarial network to obtain the output reconstructed sequencing dataset.

[0015] The platform label set output module is used to determine, through the discriminator in the generative adversarial network, a first platform label set corresponding to the transcription sequencing training set and a second platform label set corresponding to the reconstructed sequencing dataset;

[0016] A generative adversarial network training module is used to perform alternating iterative training on the generator and the discriminator based on the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set to obtain a trained generative adversarial network.

[0017] The generator in the trained generative adversarial network is used to reconstruct the transcriptional sequencing data to be tested from any sequencing platform.

[0018] According to another embodiment of the present invention, an electronic device is provided, the electronic device comprising:

[0019] At least one processor; and

[0020] A memory communicatively connected to the at least one processor; wherein,

[0021] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the transcription sequencing data processing method according to any embodiment of the present invention.

[0022] According to another embodiment of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the transcription sequencing data processing method described in any embodiment of the present invention.

[0023] According to another embodiment of the present invention, a computer program product is provided, including a computer program that, when executed by a processor, implements the transcription sequencing data processing method described in any embodiment of the present invention.

[0024] The technical solution of this embodiment trains a generative adversarial network (GAN) using transcriptional sequencing training sets from at least two sequencing platforms. The generator in the GAN learns adversarially to map transcriptional sequencing data from different sequencing platforms to a unified vector space, achieving data distribution alignment between sequencing platforms. This solves the problem of poor platform differentiation removal in traditional methods. The discriminator in the GAN identifies platforms on both the transcriptional sequencing training set and the reconstructed sequencing dataset, reducing the ability to identify platforms on the reconstructed sequencing dataset while retaining the ability to identify platforms on the transcriptional sequencing training set. This eliminates cross-platform sequencing bias while preserving the biological characteristics contained in the transcriptional sequencing data.

[0025] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 A flowchart illustrating a method for processing transcription sequencing data according to an embodiment of the present invention;

[0028] Figure 2 A network architecture diagram of a specific instance of a generative adversarial network provided in an embodiment of the present invention;

[0029] Figure 3 A flowchart illustrating another method for processing transcription sequencing data provided in one embodiment of the present invention;

[0030] Figure 4 A flowchart illustrating another method for processing transcription sequencing data provided in one embodiment of the present invention;

[0031] Figure 5 A flowchart illustrating a specific example of a method for processing transcription sequencing data provided in an embodiment of the present invention;

[0032] Figure 6 This is a schematic diagram of a transcription sequencing data processing device provided in one embodiment of the present invention;

[0033] Figure 7 This is a schematic diagram of the structure of an electronic device provided in one embodiment of the present invention. Detailed Implementation

[0034] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0035] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0036] Figure 1 This is a flowchart of a method for processing transcription sequencing data according to an embodiment of the present invention. This embodiment is applicable to the case of de-platformization of transcription sequencing data. The method can be executed by a transcription sequencing data processing device, which can be implemented in hardware and / or software and can be configured in a terminal device.

[0037] like Figure 1 As shown, the method includes:

[0038] S110. Obtain transcription sequencing training sets from at least two sequencing platforms.

[0039] Specifically, a sequencing platform refers to a technical system used to determine the base sequence of RNA (ribonucleic acid) molecules. Its core components include a sequencer, chemical reagents, signal detection, and data analysis. For example, the sequencing technology used in a sequencing platform may be sequencing-while-synthesizing, single-molecule real-time sequencing, or nanopore electrophysiological sequencing, but it is not limited to the examples described above.

[0040] Specifically, the transcription sequencing training set contains multiple transcription sequencing data for training. The transcription sequencing data characterizes the quantitative information obtained by the sequencing platform from RNA sequencing of the sequencing samples. The transcription sequencing data contains gene expression values ​​corresponding to multiple sequencing genes, and the gene expression values ​​can be used to measure gene transcription activity.

[0041] In one optional embodiment, when the sequencing sample is a single-cell sample, the transcription sequencing data in the transcription sequencing training set is single-cell transcription sequencing data; when the sequencing sample is a tissue sample or a cell population sample, the transcription sequencing data in the transcription sequencing training set is batch transcription sequencing data.

[0042] S120. Input the transcription sequencing training set into the generator in the generative adversarial network to obtain the output reconstructed sequencing dataset.

[0043] Generative Adversarial Network (GAN) is a deep learning framework consisting of a generator and a discriminator. Through adversarial training, the generator learns to generate platform-free reconstructed sequencing data, while the discriminator learns to distinguish the sequencing platform to which the transcribed sequencing data in the transcribed sequencing training set belongs, as well as the sequencing platform to which the reconstructed sequencing data in the reconstructed sequencing dataset belongs.

[0044] Specifically, each transcriptional sequencing data in the transcriptional sequencing training set is input into the generator in the generative adversarial network to obtain the reconstructed sequencing data corresponding to each transcriptional sequencing data. Multiple reconstructed sequencing data constitute the reconstructed sequencing dataset.

[0045] In an optional embodiment, the generator is a variational autoencoder, which includes an encoder, a hidden layer, and a decoder. The encoder is used to encode the input transcription sequencing training set to obtain an encoded sequencing dataset. The hidden layer is used to reduce the dimensionality of the encoded sequencing dataset to obtain a dimensionality-reduced sequencing dataset. The decoder is used to decode the dimensionality-reduced sequencing dataset to obtain an output reconstructed sequencing dataset.

[0046] S130. By generating a discriminator in the adversarial network, determine the first platform label set corresponding to the transcription sequencing training set and the second platform label set corresponding to the reconstructed sequencing dataset.

[0047] Specifically, each transcribed sequencing data in the transcription sequencing training set is input into the discriminator to obtain the first platform label corresponding to each transcribed sequencing data output. Multiple first platform labels constitute the first platform label set. Each reconstructed sequencing data in the reconstructed sequencing training set is input into the discriminator to obtain the second platform label corresponding to each reconstructed sequencing data output. Multiple second platform labels constitute the second platform label set.

[0048] Specifically, the first platform label indicates which sequencing platform the transcription sequencing data originated from, and the second platform label indicates which sequencing platform the reconstructed sequencing data originated from.

[0049] Figure 2 This is a network architecture diagram of a specific instance of a generative adversarial network provided in an embodiment of the present invention, wherein boxes a, b and c represent transcriptional sequencing data in the transcriptional sequencing training set, boxes a1, b1 and c1 represent reconstructed sequencing data in the reconstructed sequencing data, and the colors represent the platform labels corresponding to the sequencing platforms.

[0050] Specifically, the Generative Adversarial Network (GAN) comprises a generator and a discriminator. The generator is a variational autoencoder (VAE), which includes an encoder, hidden layers, and a decoder. The encoder consists of a linear layer, a ReLU activation function, batch normalization, and Dropout regularization. The encoder's output includes the mean, variance, and a random number. The hidden layer performs logarithmic processing on the variance, multiplies the logarithmic result with the random number, and adds the product to the mean to obtain the latent feature vector Z. The random number is sampled from a normal distribution. The decoder includes a reverse structure symmetric to the encoder. The discriminator consists of two layers, each including a linear layer, a LeakyReLU activation function, and Dropout regularization.

[0051] The embodiments of the present invention only provide illustrative explanations of the network architecture of generative adversarial networks and do not limit it. It is understood that in other embodiments, attention mechanisms or graph neural networks may be introduced into generative adversarial networks to further improve their learning and generalization capabilities.

[0052] S140. Based on the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set, the generator and discriminator are trained iteratively to obtain the trained generative adversarial network.

[0053] Specifically, during the alternating iterative training process, the discriminator is fixed, and the generator is optimized based on the function values ​​determined by the transcription sequencing training set and the reconstructed sequencing dataset to eliminate technical noise between different reconstructed sequencing data; the generator is fixed, and the discriminator is optimized based on the loss function values ​​determined by the first platform label set and the second platform label set to make the platform label distribution corresponding to the first platform label set close to the real platform label distribution, and the platform label distribution corresponding to the second platform label set close to a uniform distribution.

[0054] In this embodiment, the generator in the trained generative adversarial network is used to reconstruct the transcriptional sequencing data to be tested from any sequencing platform.

[0055] The technical solution of this embodiment trains a generative adversarial network (GAN) using transcriptional sequencing training sets from at least two sequencing platforms. The generator in the GAN learns adversarially to map transcriptional sequencing data from different sequencing platforms to a unified vector space, achieving data distribution alignment between sequencing platforms. This solves the problem of poor platform differentiation removal in traditional methods. The discriminator in the GAN identifies platforms on both the transcriptional sequencing training set and the reconstructed sequencing dataset, reducing the ability to identify platforms on the reconstructed sequencing dataset while retaining the ability to identify platforms on the transcriptional sequencing training set. This eliminates cross-platform sequencing bias while preserving the biological characteristics contained in the transcriptional sequencing data.

[0056] Figure 3 This is a flowchart of another method for processing transcription sequencing data according to an embodiment of the present invention. This embodiment further refines the "obtaining transcription sequencing training sets from at least two sequencing platforms" in the above embodiment. In this embodiment, the transcription sequencing data is batch transcription sequencing data. Obtaining transcription sequencing training sets from at least two sequencing platforms includes: obtaining single-cell sequencing datasets corresponding to at least two sequencing platforms respectively; for each sequencing platform, according to the cell mixing ratio, fusing tumor sequencing data and immune sequencing data in the single-cell sequencing dataset corresponding to the sequencing platform to obtain pseudo-batch transcription sequencing data; and constructing a transcription sequencing training set based on the pseudo-batch transcription sequencing data corresponding to each sequencing platform.

[0057] like Figure 3 As shown, the method includes:

[0058] S210. Obtain single-cell sequencing datasets corresponding to at least two sequencing platforms.

[0059] In this embodiment, the single-cell sequencing dataset includes tumor sequencing data corresponding to at least one tumor cell line and immune sequencing data corresponding to at least one immune cell.

[0060] For example, tumor cell lines can be rectal cancer cell lines HCT116, HT29, DLD1, SW1463, CRC#21, lung cancer cell lines, liver cancer cell lines, and breast cancer cell lines, etc., but are not limited to the examples above.

[0061] In this embodiment, the single-cell sequencing dataset includes immune sequencing data corresponding to B cells, immune sequencing data corresponding to T cells, and immune sequencing data corresponding to macrophages.

[0062] S220. For each sequencing platform, based on the cell mixing ratio, the tumor sequencing data and immune sequencing data in the single-cell sequencing dataset corresponding to the sequencing platform are fused to obtain pseudo-batch transcription sequencing data.

[0063] For example, for each sequencing platform, pseudo-batch transcription sequencing data can be obtained by fusing tumor sequencing data of the corresponding rectal cancer cell line HCT116, immune sequencing data of B cells, immune sequencing data of T cells, and immune sequencing data of macrophages, according to the cell mixing ratio. Pseudo-batch transcription sequencing data can be shown in Table 1.

[0064] Table 1: Pseudo-batch transcription sequencing data

[0065]

[0066] Specifically, the cell mixing ratio includes the mixing ratio of tumor cell lines and each immune cell.

[0067] In an optional embodiment, the method further includes: for each tumor cell line, determining a gene expression matrix based on the tumor immune data corresponding to the tumor cell line and the immune sequencing data corresponding to at least one immune cell in the single-cell sequencing dataset; and performing deconvolution processing on the gene expression matrix to obtain the cell mixing ratio.

[0068] In this embodiment, the matrix parameters in the gene expression matrix represent the expression values ​​of sequenced genes in tumor sequencing data or immune sequencing data.

[0069] In another optional embodiment, the method further includes: for each tumor cell line, randomly generating a tumor mixing ratio corresponding to the tumor cell line according to a first ratio range; randomly generating a first mixing ratio corresponding to T cells according to a second ratio range, and determining a second mixing ratio corresponding to B cells and a third mixing ratio corresponding to macrophages according to the tumor mixing ratio and the first mixing ratio; and determining a cell mixing ratio corresponding to the tumor cell line according to the tumor mixing ratio, the first mixing ratio, the second mixing ratio, and the third mixing ratio.

[0070] In this embodiment, the first proportional range and the second proportional range do not overlap. For example, the first proportional range can be [0.1, 0.6] and the second proportional range can be [0.7, 0.9], but it is not limited to the example scenario.

[0071] The sum of the proportions of the tumor mixing ratio, the first mixing ratio, the second mixing ratio, and the third mixing ratio can be 1.

[0072] In one optional embodiment, based on the cell mixing ratio, the tumor sequencing data and immune sequencing data in the single-cell sequencing dataset corresponding to the sequencing platform are fused to obtain pseudo-batch transcription sequencing data, including: based on the cell mixing ratio, the tumor sequencing data and immune sequencing data in the single-cell sequencing dataset corresponding to the sequencing platform are weighted summed or weighted averaged according to the expression values ​​to obtain pseudo-batch transcription sequencing data.

[0073] In another optional embodiment, based on the cell mixing ratio, the tumor sequencing data and immune sequencing data in the single-cell sequencing dataset corresponding to the sequencing platform are fused to obtain pseudo-batch transcription sequencing data. This includes: determining a transcription sequencing matrix based on the tumor sequencing data with the tumor mixing ratio in the single-cell sequencing dataset, the immune sequencing data corresponding to T cells with the first mixing ratio, the immune sequencing data corresponding to B cells with the second mixing ratio, and the immune sequencing data corresponding to macrophages with the third mixing ratio; and performing convolution processing on the transcription sequencing matrix to obtain pseudo-batch transcription sequencing data corresponding to the tumor cell line.

[0074] For example, pseudo-batch transcription sequencing data corresponding to tumor cell lines can be shown in Table 2.

[0075] Convolution is a binary mathematical operation defined on two functions (input function f and kernel function h). Its core operations include flipping, translation (sliding), and superposition (summing). In this embodiment, f can represent the transcription sequencing matrix, and h can be a kernel function used for feature extraction or signal processing.

[0076] Table 2: Pseudo-batch transcription sequencing data corresponding to tumor cell lines

[0077]

[0078] In the continuous domain, convolution is represented as the integral of the product of two functions, while in the discrete domain, it is represented as a weighted sum of two functions. The convolution process mainly involves two basic steps: flipping and sliding stacking.

[0079] First, the kernel function can be... Perform a mirror flip, and then perform the flipped kernel function. Along the input function The domain is shifted, and the weighted sum of each pair of functions is calculated point by point in the discrete domain.

[0080] For example, convolution processing can be represented by the following formula:

[0081] ;

[0082] in, It can be the result of a convolution operation, i.e., pseudo-batch transcription and sequencing data, specifically a two-dimensional matrix. Represents the transcription sequencing matrix. Represents the kernel function. This represents the matrix coordinates used to traverse the transcription sequencing matrix. Indicates the location Input value at the location, This indicates that after flipping the kernel function h, at position... The value at that location.

[0083] The advantage of setting up convolutional fusion is that convolutional operations, through local receptive fields and weight sharing mechanisms, can effectively capture the spatial correlation of specific gene expression in different cell types. This more accurately simulates the nonlinear transcriptional features generated by cell population interactions in real batch transcription sequencing data, improves the similarity between pseudo-batch transcription sequencing data and real batch transcription sequencing data, and shows that pseudo-batch transcription sequencing data exhibits high stability and biological consistency.

[0084] S230. Construct a transcription sequencing training set based on the pseudo-batch transcription sequencing data corresponding to each sequencing platform.

[0085] Specifically, the transcription sequencing training set contains pseudo-batch transcription sequencing data corresponding to each sequencing platform, or the transcription sequencing training set contains pseudo-batch transcription sequencing data and real batch transcription sequencing data corresponding to each sequencing platform.

[0086] S240. Input the transcription sequencing training set into the generator in the generative adversarial network to obtain the output reconstructed sequencing dataset.

[0087] S250. By generating a discriminator in a generative adversarial network, determine the first platform label set corresponding to the transcription sequencing training set and the second platform label set corresponding to the reconstructed sequencing dataset.

[0088] S260. Based on the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set, the generator and discriminator are trained iteratively to obtain the trained generative adversarial network.

[0089] S240-S260 in this embodiment are the same as those in the above embodiments. Figure 1 The S120-S140 shown are the same or similar, and will not be described again in this embodiment.

[0090] The technical solution of this embodiment obtains single-cell sequencing datasets corresponding to at least two sequencing platforms. For each sequencing platform, based on the cell mixing ratio, the tumor sequencing data and immune sequencing data in the single-cell sequencing dataset corresponding to the sequencing platform are fused to obtain pseudo-batch transcription sequencing data. Based on the pseudo-batch transcription sequencing data corresponding to each sequencing platform, a transcription sequencing training set is constructed. This solves the problem of limited data volume of batch transcription sequencing data for tumor tissue samples, alleviates overfitting caused by the scarcity of real batch transcription sequencing data, and reduces batch effects related to biological differences and experimental procedures. Especially in cross-platform integration tasks, pseudo-batch transcription sequencing data can cover a wider range of technical bias scenarios and improve the generalization ability of generative adversarial networks.

[0091] Figure 4 This is a flowchart of another method for processing transcriptional sequencing data according to an embodiment of the present invention. This embodiment further refines the step in the above embodiment of "alternating iterative training of the generator and discriminator to obtain a trained generative adversarial network based on the transcriptional sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set." In this embodiment, alternating iterative training of the generator and discriminator to obtain a trained generative adversarial network based on the transcriptional sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set includes: determining a first loss function value based on the transcriptional sequencing training set and the reconstructed sequencing dataset, and adjusting the generator's model parameters based on the first loss function value while keeping the discriminator fixed; determining a second loss function value based on the first platform label set and the second platform label set, and adjusting the discriminator's model parameters based on the second loss function value while keeping the generator fixed; until both the first loss function value and the second loss function value converge, a trained generative adversarial network is obtained.

[0092] like Figure 4 As shown, the method includes:

[0093] S310. Obtain transcription sequencing training sets from at least two sequencing platforms.

[0094] S320. Input the transcription sequencing training set into the generator in the generative adversarial network to obtain the output reconstructed sequencing dataset.

[0095] S330. By generating a discriminator in the adversarial network, determine the first platform label set corresponding to the transcription sequencing training set and the second platform label set corresponding to the reconstructed sequencing dataset.

[0096] S310-S330 in this embodiment are the same as those in the above embodiments. Figure 1The S110-S130 shown are the same as or similar to those in the above embodiments. Figure 2 The S210-S250 shown are the same or similar, and will not be described again in this embodiment.

[0097] S340. Based on the transcription sequencing training set and the reconstructed sequencing dataset, determine the first loss function value, and adjust the generator model parameters according to the first loss function value while keeping the discriminator fixed.

[0098] In an optional embodiment, the first loss function value characterizes the reconstruction loss of the generator. For example, the loss function corresponding to the first loss function value is the mean squared error function, and the first loss function value satisfies the following formula:

[0099] ;

[0100] in, This indicates the number of transcriptional sequencing data in the transcriptional sequencing training set. Indicates the first One transcriptional sequencing data, Indicates the first One reconstructed sequencing data.

[0101] In another optional embodiment, the generator's output also includes a predicted classification label set, wherein the predicted classification labels in the predicted classification label set correspond one-to-one with the transcriptional sequencing data in the transcriptional sequencing training set. Accordingly, based on the transcriptional sequencing training set and the reconstructed sequencing dataset, a first loss function value is determined, including: determining a third loss function value based on the transcriptional sequencing training set and the reconstructed sequencing dataset, and determining a fourth loss function value based on the predicted classification label set and the standard classification label set corresponding to the transcriptional sequencing training set; and determining a first loss function value based on the third loss function value and the fourth loss function value.

[0102] In this embodiment, the third loss function value represents the reconstruction loss of the generator, and the fourth loss function value represents the label distribution loss of the generator, which is used to measure the difference between the latent distribution and the prior distribution.

[0103] For example, the loss function corresponding to the fourth loss function value is the KL (Kullback-Leibler) divergence function, and the fourth loss function value satisfies the following formula:

[0104] ;

[0105] in, Indicates the first in the standard classification label set Probability distribution data corresponding to each predicted class label Indicates the first in the predicted category label set The probability distribution data corresponding to each predicted category label.

[0106] For example, the first loss function value

[0107] It should be noted that when determining the value of the first loss function based on the transcription sequencing training set and the reconstructed sequencing dataset, This is the first loss function value. When the third loss function value is determined based on the transcription sequencing training set and the reconstructed sequencing dataset, and the fourth loss function value is determined based on the predicted classification label set and the standard classification label set corresponding to the transcription sequencing training set; when determining the first loss function value based on the third and fourth loss function values, This can be the third loss function value; in this case, the first loss function value can be... .

[0108] S350. Based on the first platform label set and the second platform label set, determine the second loss function value, and with the generator fixed, adjust the model parameters of the discriminator based on the second loss function value.

[0109] In an optional embodiment, determining the second loss function value based on the first platform label set and the second platform label set includes: determining the fifth loss function value based on the first platform label set and determining the sixth loss function value based on the second platform label set; and determining the second loss function value based on the fifth loss function value and the sixth loss function value.

[0110] In this embodiment, the optimization objective of the discriminator is to make the platform label distribution corresponding to the first platform label set close to the real platform label distribution, and the platform label distribution corresponding to the second platform label set close to a uniform distribution.

[0111] For example, the fifth and sixth loss function values ​​can both be based on the cross-entropy loss function ( ) Construction.

[0112] Second loss function value Satisfy the following formula:

[0113] ;

[0114] in, This represents the value of the fifth loss function. This represents the value of the sixth loss function.

[0115] For example, the crossover loss function can be expressed by the following formula:

[0116] ;

[0117] in, Cross-entropy loss can be used to measure the difference between the predicted distribution and the true distribution output by the discriminator. The smaller the cross-entropy loss value, the more accurate the prediction. Indicates the first The true probability distribution corresponding to each sequencing platform label Indicates the middle Predicted probability distributions corresponding to each sequencing platform tag.

[0118] Among them, when When representing the fifth loss function, This indicates that the first platform tag set, the... The true probability distribution corresponding to each sequencing platform label This indicates that the first platform tag set, the... Predicted probability distributions corresponding to each sequencing platform tag.

[0119] when When representing the sixth loss function, This indicates that the second platform tag set is being referred to as the first. The true probability distribution corresponding to each sequencing platform label This indicates that the second platform tag set is being referred to as the first. Predicted probability distributions corresponding to each sequencing platform tag.

[0120] S360, until both the first and second loss function values ​​converge, the trained generative adversarial network is obtained.

[0121] The technical solution of this embodiment sets the optimization target of the discriminator to be close to the real platform label distribution corresponding to the first platform label set, and close to the uniform distribution of the platform label distribution corresponding to the second platform label set. This not only preserves the discriminator's ability to identify platform features, but also ensures that the reconstructed sequencing data output by the generator eliminates platform specificity as much as possible, thereby improving the elimination effect of cross-platform sequencing bias.

[0122] Based on the above embodiments, the method may optionally further include: acquiring test transcription sequencing data corresponding to multiple sequencing platforms respectively, and inputting the multiple test transcription sequencing data into the generator in the trained generative adversarial network to obtain the reconstructed sequencing data corresponding to each test transcription sequencing data; inputting the multiple reconstructed sequencing data into a classification and discrimination model to obtain the predicted discrimination label corresponding to each output transcription sequencing data.

[0123] For example, the predicted discriminant label can be a cancer category, such as lung cancer or liver cancer. The predicted discriminant label can also be a tumor subtype, such as histological subtype, molecular subtype, driver gene subtype, MSI status (Microsatellite Instability), etc. The predicted discriminant label can also be a survival category, such as survival probability, but it is not limited to the above examples.

[0124] Based on the above embodiments, optionally, the method further includes: obtaining the reconstructed sequencing set corresponding to each predicted discriminant label; for the reconstructed sequencing data in each reconstructed sequencing set, obtaining the co-cluster distances between the reconstructed sequencing data and the remaining reconstructed sequencing data in the reconstructed sequencing set, and obtaining the hetero-cluster distances between the reconstructed sequencing data and the reconstructed sequencing data in other reconstructed sequencing sets; determining the contour coefficients corresponding to the reconstructed sequencing data based on the multiple co-cluster distances and the multiple hetero-cluster distances; and determining the average contour coefficient based on the contour coefficients corresponding to the multiple reconstructed sequencing data.

[0125] The mean silhouette coefficient is an indicator used to measure the quality of clustering. The larger the mean silhouette coefficient, the more biological feature information is retained in the reconstructed sequencing data. Conversely, the smaller the mean silhouette coefficient, the less biological feature information is retained in the reconstructed sequencing data.

[0126] Based on the above embodiments, the method may optionally further include: obtaining a reconstructed sequencing set corresponding to any one of the predicted discriminant labels, and determining a first transcriptional sequencing dataset corresponding to the reconstructed sequencing set; obtaining a second transcriptional sequencing dataset corresponding to the predicted discriminant label; and determining neighborhood consistency based on the first transcriptional sequencing dataset and the second transcriptional sequencing dataset.

[0127] In this embodiment, for example, the second transcription sequencing dataset is obtained by clustering multiple transcription sequencing data or by classifying and predicting multiple transcription sequencing data through a neural network model.

[0128] Specifically, neighborhood consistency is used to measure the fidelity of local structure. The greater the neighborhood consistency, the better the effect of removing plateau bias from the reconstructed sequencing data. Conversely, the smaller the neighborhood consistency, the worse the effect of removing plateau bias from the reconstructed sequencing data.

[0129] Based on the above embodiments, optionally, the method further includes: obtaining reconstructed sequencing data corresponding to the output test transcription sequencing data from the generator in a generative adversarial network trained with test transcription sequencing data from a public database.

[0130] Specifically, test transcription sequencing data from public databases were used as an independent test set to verify the generalization performance of generative adversarial networks.

[0131] Based on the above embodiments, the method may optionally further include: determining the platform integration score based on the average profile coefficient and neighborhood consistency.

[0132] Using the transcription sequencing data processing method provided in the above embodiments of the present invention, 2512 batch transcription sequencing data from multiple sequencing platforms were tested, and the platform integration score was 0.757±0.014.

[0133] Compared to the traditional Limma model, the platform integration score is improved by 10%-18%; compared to the traditional Combat model, the platform integration score is improved by 21%-28%; and compared to the traditional Flow model, the platform integration score is improved by 8%-15%. Comparing only neighborhood consistency, the embodiments of the present invention improve neighborhood consistency by 16%-36%, 16%-25%, and 10%-19% respectively compared to the traditional Limma, Combat, and Flow models.

[0134] The test data above show that the embodiments of the present invention effectively improve the accuracy of downstream classification tasks by preprocessing batch transcription sequencing data from different sequencing platforms.

[0135] Figure 5 This is a flowchart illustrating a specific example of a method for processing transcriptional sequencing data according to an embodiment of the present invention. Specifically, in the data preparation stage, single-cell sequencing data are collected from multiple sequencing platforms, and tumor sequencing data and immune sequencing data in the single-cell sequencing dataset are fused according to the cell mixing ratio to obtain pseudo-batch transcriptional sequencing data simulating tumor tissue. In the model training stage, the generator and discriminator in the generative adversarial network are iteratively trained alternately based on the transcriptional sequencing training set constructed from the pseudo-batch transcriptional sequencing data to obtain the trained generative adversarial network. Specifically, the mean squared error loss function and the KL loss function are used to determine the first loss function value for training the generator, and the cross-entropy loss function is used to determine the second loss function value for training the discriminator.

[0136] In the model application phase, the transcriptional sequencing data from multiple sequencing platforms are input into the generator in the trained generative adversarial network to obtain the output reconstructed sequencing data. Downstream tasks, such as cancer classification, tumor subtype discrimination, and survival classification, are then performed on these reconstructed sequencing data. In the method validation phase, generalization validation is performed using transcriptional sequencing data from public databases. Based on the task data corresponding to the downstream tasks, metrics such as average contour width and neighborhood consistency are calculated, and the results are visualized.

[0137] The following are embodiments of the transcription sequencing data processing device provided in this invention. This device and the transcription sequencing data processing method in the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the transcription sequencing data processing device, please refer to the content of the transcription sequencing data processing method in the above embodiments.

[0138] Figure 6 This is a schematic diagram of a transcription sequencing data processing device provided in one embodiment of the present invention. Figure 6 As shown, the device includes: a transcription sequencing training set acquisition module 410, a reconstructed sequencing dataset output module 420, a platform label set output module 430, and a generative adversarial network training module 440.

[0139] The system includes: a transcription sequencing training set acquisition module 410, used to acquire transcription sequencing training sets from at least two sequencing platforms; a reconstructed sequencing dataset output module 420, used to input the transcription sequencing training sets into the generator in the generative adversarial network (GAN) to obtain the output reconstructed sequencing dataset; a platform label set output module 430, used to determine the first platform label set corresponding to the transcription sequencing training set and the second platform label set corresponding to the reconstructed sequencing dataset through the discriminator in the GAN; and a GAN training module 440, used to iteratively train the generator and discriminator based on the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set to obtain the trained GAN.

[0140] The generator in the trained generative adversarial network is used to reconstruct the transcriptional sequencing data to be tested from any sequencing platform.

[0141] The technical solution of this embodiment trains a generative adversarial network (GAN) using transcriptional sequencing training sets from at least two sequencing platforms. The generator in the GAN learns adversarially to map transcriptional sequencing data from different sequencing platforms to a unified vector space, achieving data distribution alignment between sequencing platforms. This solves the problem of poor platform differentiation removal in traditional methods. The discriminator in the GAN identifies platforms on both the transcriptional sequencing training set and the reconstructed sequencing dataset, reducing the ability to identify platforms on the reconstructed sequencing dataset while retaining the ability to identify platforms on the transcriptional sequencing training set. This eliminates cross-platform sequencing bias while preserving the biological characteristics contained in the transcriptional sequencing data.

[0142] In an optional embodiment, the transcription sequencing training set acquisition module 410 includes:

[0143] The single-cell sequencing dataset acquisition unit is used to acquire single-cell sequencing datasets corresponding to at least two sequencing platforms when the transcription sequencing data is batch transcription sequencing data; wherein, the single-cell sequencing dataset contains tumor sequencing data corresponding to at least one tumor cell line and immune sequencing data corresponding to at least one immune cell.

[0144] The pseudo-batch transcription sequencing data determination unit is used to fuse tumor sequencing data and immune sequencing data in the single-cell sequencing dataset corresponding to each sequencing platform according to the cell mixing ratio to obtain pseudo-batch transcription sequencing data.

[0145] The transcription sequencing training set construction unit is used to construct a transcription sequencing training set based on the pseudo-batch transcription sequencing data corresponding to each sequencing platform.

[0146] In an optional embodiment, the device further includes:

[0147] The cell mixing ratio determination module is used to randomly generate the tumor mixing ratio corresponding to each tumor cell line based on a first ratio range.

[0148] Based on the second ratio range, a first mixing ratio corresponding to T cells is randomly generated, and based on the tumor mixing ratio and the first mixing ratio, a second mixing ratio corresponding to B cells and a third mixing ratio corresponding to macrophages are determined.

[0149] The cell mixing ratio corresponding to the tumor cell line is determined based on the tumor mixing ratio, the first mixing ratio, the second mixing ratio, and the third mixing ratio.

[0150] The first proportional range and the second proportional range do not overlap.

[0151] In one optional embodiment, the pseudo-batch transcription sequencing data determination unit is specifically used for:

[0152] The transcription sequencing matrix is ​​determined based on the tumor sequencing data of the mixed proportion of tumors in the single-cell sequencing dataset, the immune sequencing data corresponding to the first mixed proportion of T cells, the immune sequencing data corresponding to the second mixed proportion of B cells, and the immune sequencing data corresponding to the third mixed proportion of macrophages.

[0153] The transcription sequencing matrix was convolved to obtain pseudo-batch transcription sequencing data corresponding to the tumor cell line.

[0154] In an optional embodiment, the generator is a variational autoencoder, which includes an encoder, a hidden layer, and a decoder. The encoder is used to encode the input transcription sequencing training set to obtain an encoded sequencing dataset. The hidden layer is used to reduce the dimensionality of the encoded sequencing dataset to obtain a dimensionality-reduced sequencing dataset. The decoder is used to decode the dimensionality-reduced sequencing dataset to obtain an output reconstructed sequencing dataset.

[0155] In an optional embodiment, the adversarial network training module 440 includes:

[0156] The first loss function value determination unit is used to determine the first loss function value based on the transcription sequencing training set and the reconstructed sequencing dataset, and adjust the generator model parameters based on the first loss function value while keeping the discriminator fixed. The second loss function value determination unit is used to determine the second loss function value based on the first platform label set and the second platform label set, and adjust the discriminator model parameters based on the second loss function value while keeping the generator fixed. The generative adversarial network determination unit is used to obtain the trained generative adversarial network until both the first loss function value and the second loss function value converge.

[0157] In an optional embodiment, the generator's output further includes a predicted classification label set, wherein the predicted classification labels in the predicted classification label set correspond one-to-one with the transcriptional sequencing data in the transcriptional sequencing training set. Accordingly, the first loss function value determination unit is specifically used to: determine a third loss function value based on the transcriptional sequencing training set and the reconstructed sequencing dataset, and determine a fourth loss function value based on the predicted classification label set and the standard classification label set corresponding to the transcriptional sequencing training set; and determine a first loss function value based on the third loss function value and the fourth loss function value.

[0158] In an optional embodiment, the second loss function value determination unit is specifically configured to: determine a fifth loss function value based on a first platform tag set, and determine a sixth loss function value based on a second platform tag set; and determine a second loss function value based on the fifth loss function value and the sixth loss function value.

[0159] The optimization objective of the discriminator is to make the platform label distribution corresponding to the first platform label set close to the real platform label distribution, and the platform label distribution corresponding to the second platform label set close to a uniform distribution.

[0160] The transcription sequencing data processing apparatus provided in the embodiments of the present invention can execute the transcription sequencing data processing method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0161] Figure 7This is a schematic diagram of an electronic device provided according to one embodiment of the present invention. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0162] like Figure 7 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor 11. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0163] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information or data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0164] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the transcription sequencing data processing method provided in the above embodiments.

[0165] In some embodiments, the transcription sequencing data processing method provided in the above embodiments can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the transcription sequencing data processing method described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the transcription sequencing data processing method by any other suitable means (e.g., by means of firmware).

[0166] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication unit 19, or installed from storage unit 18, or installed from ROM 12. When the computer program is executed by processor 11, it performs the functions defined in the methods of the embodiments of the present invention.

[0167] Various embodiments of the systems and techniques described above herein can be implemented in the following systems or combinations thereof: digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard parts (ASSPs), system-on-chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0168] Computer programs for implementing the transcription sequencing data processing methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to the processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer programs can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0169] In the context of this application, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium can be a machine-readable storage medium. Examples of machine-readable storage media include, based on an electrical connection of at least one wire, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0170] To provide interaction with a user, the systems and techniques described herein can be implemented on a terminal device having: a display device for displaying information to the user (e.g., a cathode-ray tube (CRT) or liquid crystal display (LCD) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the terminal device. Other types of devices can also provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0171] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0172] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system. It addresses the shortcomings of traditional physical hosts and Virtual Private Server (VPS) services, such as high management difficulty and weak business scalability.

[0173] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0174] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for processing transcription sequencing data, characterized in that, include: Obtain transcription sequencing training sets from at least two sequencing platforms; The transcriptional sequencing training set is input into the generator in the generative adversarial network to obtain the output reconstructed sequencing dataset; The discriminator in the generative adversarial network determines a first platform label set corresponding to the transcription sequencing training set and a second platform label set corresponding to the reconstructed sequencing dataset. Based on the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set, the generator and the discriminator are alternately iteratively trained to obtain a trained generative adversarial network; The generator in the trained generative adversarial network is used to reconstruct the transcriptional sequencing data to be tested from any sequencing platform.

2. The method for processing transcription sequencing data according to claim 1, characterized in that, The acquisition of transcription sequencing training sets from at least two sequencing platforms includes: When the transcription sequencing data is batch transcription sequencing data, at least two single-cell sequencing datasets corresponding to different sequencing platforms are obtained; wherein, the single-cell sequencing datasets contain tumor sequencing data corresponding to at least one tumor cell line and immune sequencing data corresponding to at least one immune cell. For each sequencing platform, based on the cell mixing ratio, the tumor sequencing data and immune sequencing data in the single-cell sequencing dataset corresponding to the sequencing platform are fused to obtain pseudo-batch transcription sequencing data. A transcription sequencing training set is constructed based on the pseudo-batch transcription sequencing data corresponding to each sequencing platform.

3. The method for processing transcription sequencing data according to claim 2, characterized in that, Also includes: For each tumor cell line, a tumor mixing ratio corresponding to the tumor cell line is randomly generated according to a first ratio range; Based on the second ratio range, a first mixing ratio corresponding to T cells is randomly generated, and based on the tumor mixing ratio and the first mixing ratio, a second mixing ratio corresponding to B cells and a third mixing ratio corresponding to macrophages are determined. The cell mixing ratio corresponding to the tumor cell line is determined based on the tumor mixing ratio, the first mixing ratio, the second mixing ratio, and the third mixing ratio. Wherein, the first ratio range and the second ratio range do not overlap.

4. The method for processing transcription sequencing data according to claim 3, characterized in that, The process of fusing tumor sequencing data and immune sequencing data from the single-cell sequencing dataset corresponding to the sequencing platform according to the cell mixing ratio to obtain pseudo-batch transcription sequencing data includes: The transcription sequencing matrix is ​​determined based on the tumor sequencing data of the mixed proportion of tumors in the single-cell sequencing dataset, the immune sequencing data corresponding to the first mixed proportion of T cells, the immune sequencing data corresponding to the second mixed proportion of B cells, and the immune sequencing data corresponding to the third mixed proportion of macrophages. The transcription sequencing matrix is ​​convolved to obtain pseudo-batch transcription sequencing data corresponding to the tumor cell line.

5. The method for processing transcription sequencing data according to claim 1 or 2, characterized in that, The generator is a variational autoencoder, which includes an encoder, a hidden layer, and a decoder. The encoder is used to encode the input transcription sequencing training set to obtain an encoded sequencing dataset. The hidden layer is used to reduce the dimensionality of the encoded sequencing dataset to obtain a dimensionality-reduced sequencing dataset. The decoder is used to decode the dimensionality-reduced sequencing dataset to obtain an output reconstructed sequencing dataset.

6. The method for processing transcription sequencing data according to claim 1 or 2, characterized in that, The step of iteratively training the generator and the discriminator using the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set to obtain a trained generative adversarial network includes: Based on the transcription sequencing training set and the reconstructed sequencing dataset, a first loss function value is determined, and with the discriminator fixed, the model parameters of the generator are adjusted according to the first loss function value; Based on the first platform label set and the second platform label set, a second loss function value is determined, and with the generator fixed, the model parameters of the discriminator are adjusted according to the second loss function value; The training of the generative adversarial network is completed when both the first loss function value and the second loss function value converge.

7. The method for processing transcription sequencing data according to claim 6, characterized in that, The generator's output also includes a predicted classification label set, where each predicted classification label corresponds one-to-one with the transcriptional sequencing data in the transcriptional sequencing training set. Correspondingly, determining the first loss function value based on the transcriptional sequencing training set and the reconstructed sequencing dataset includes: Based on the transcription sequencing training set and the reconstructed sequencing dataset, a third loss function value is determined, and based on the predicted classification label set and the standard classification label set corresponding to the transcription sequencing training set, a fourth loss function value is determined. The first loss function value is determined based on the third loss function value and the fourth loss function value.

8. The method for processing transcription sequencing data according to claim 6, characterized in that, The step of determining the second loss function value based on the first platform label set and the second platform label set includes: Based on the first platform label set, determine the fifth loss function value, and based on the second platform label set, determine the sixth loss function value; The second loss function value is determined based on the fifth loss function value and the sixth loss function value; The optimization objective of the discriminator is that the platform label distribution corresponding to the first platform label set is close to the true platform label distribution, and the platform label distribution corresponding to the second platform label set is close to a uniform distribution.

9. A device for processing transcription sequencing data, characterized in that, include: The transcription sequencing training set acquisition module is used to acquire transcription sequencing training sets from at least two sequencing platforms; The reconstructed sequencing dataset output module is used to input the transcription sequencing training set into the generator in the generative adversarial network to obtain the output reconstructed sequencing dataset. The platform label set output module is used to determine, through the discriminator in the generative adversarial network, a first platform label set corresponding to the transcription sequencing training set and a second platform label set corresponding to the reconstructed sequencing dataset; A generative adversarial network training module is used to perform alternating iterative training on the generator and the discriminator based on the transcription sequencing training set, the reconstructed sequencing dataset, the first platform label set, and the second platform label set to obtain a trained generative adversarial network. The generator in the trained generative adversarial network is used to reconstruct the transcriptional sequencing data to be tested from any sequencing platform.

10. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, which enables the at least one processor to perform the method for processing transcription sequencing data according to any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the method for processing transcription sequencing data according to any one of claims 1-8.

12. A computer program product, characterized in that, The method includes a computer program that, when executed by a processor, implements the method for processing transcriptional sequencing data according to any one of claims 1-8.

Images