Prediction model training method and device, equipment and medium

By downsampling and masking processing of single-cell sequencing data, and using predictive models to correct gene expression data, the problems of insufficient sequencing depth and noise interference are solved, and the comparability of single-cell sequencing results and the accuracy of downstream tasks are improved.

CN119068985APending Publication Date: 2024-12-03BIOMAP (BEIJING) INTELLIGENCE TECH LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310630156.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-30
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The sequencing depth of single-cell sequencing results is difficult to meet expectations, resulting in incomparable results and technical noise interference, affecting the accuracy of downstream applications.

Method used

By obtaining gene expression data of multiple samples, downsampling and masking processing, gene counts at low sequencing depths are simulated, and prediction model training is used to predict gene counts at high sequencing depths, and then gene expression data is corrected to reduce noise interference.

Benefits of technology

It improves the comparability and accuracy of single-cell sequencing results, and enhances the output results of downstream tasks, especially the accuracy of cell classification, perturbation prediction and drug response prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068985A_ABST
    Figure CN119068985A_ABST
Patent Text Reader

Abstract

The invention discloses a prediction model training method and device, electronic equipment and a medium. The method comprises the steps that a plurality of samples are obtained, and each sample in the samples comprises first gene expression data, mask gene expression data and auxiliary information, the mask gene expression data is obtained by downsampling the first gene expression data and masking counting of partial genes in the downsampled first gene expression data; for each sample in the plurality of samples, processing mask gene expression data and auxiliary information in the sample by using a to-be-trained prediction model to obtain a predicted value of each gene count under the first sequencing depth corresponding to the sample; determining a loss value corresponding to the sample according to the predicted value and a count corresponding to a part of genes in the first gene expression data; and updating the to-be-trained prediction model according to the loss value corresponding to each sample in the plurality of samples.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, and in particular to a prediction model training method and device, a gene expression data correction method and device, a downstream task execution method and device, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] Single-cell sequencing technology is a technology that performs sequencing and analysis on the genome, transcriptome, and epigenome levels at the level of a single cell. Sequencing depth fundamentally affects the amount of information contained in each single-cell data and is an important factor affecting the quality of single-cell sequencing data. The deeper the sequencing depth, the lower the sparsity, and the smaller the sequencing noise, the easier it is to analyze and mine biological information from the single-cell sequencing results. Due to the influence of experimental conditions, technology, and cost, sometimes the sequencing depth of single-cell sequencing is difficult to achieve the expected level; and different single-cell sequencing results may be obtained at different sequencing depths, resulting in incomparable sequencing results obtained at different sequencing depths; in addition, even at the same sequencing depth, due to the interference of technical noise, there are certain differences in the sequencing results of similar cells. The above factors cause certain difficulties for downstream application results when single-cell sequencing results are used directly for downstream applications. Therefore, the results of single-cell sequencing need to be processed to better serve downstream applications.

[0003] The methods described in this section are not necessarily methods that have been previously conceived or employed. Unless otherwise indicated, it should not be assumed that any method described in this section is considered to be prior art simply because it is included in this section. Similarly, unless otherwise indicated, the issues mentioned in this section should not be considered to have been recognized in any prior art. Summary of the invention

[0004] According to a first aspect of the present disclosure, a method for training a prediction model is provided, comprising: obtaining a plurality of samples, wherein each of the plurality of samples comprises first gene expression data, masked gene expression data, and auxiliary information, the first gene expression data comprises counts of different genes in a single cell measured at a first sequencing depth, the masked gene expression data is obtained by downsampling the first gene expression data and masking the counts of some genes in the downsampled first gene expression data, the downsampled first gene expression data simulates the counts of different genes in a single cell measured at a second sequencing depth lower than the first sequencing depth, the auxiliary information comprises a first total count, and the first total count is the sum of the counts of each gene in the first gene expression data; for each of the plurality of samples: processing the masked gene expression data and the auxiliary information in the sample using the prediction model to be trained to obtain a predicted value of the counts of each gene at the first sequencing depth corresponding to the sample; determining a loss value corresponding to the sample according to the predicted value and the counts corresponding to some genes in the first gene expression data; and updating the prediction model to be trained according to the loss value corresponding to each of the plurality of samples.

[0005] According to a second aspect of the present disclosure, a method for correcting gene expression data is provided, comprising: obtaining current gene expression data and current auxiliary information, wherein the current gene expression data includes counts of different genes measured at an actual sequencing depth, and the current auxiliary information includes an expected first total count, the expected first total count is used to characterize the expected sequencing depth, and the expected first total count is greater than or equal to the sum of the counts of each gene in the current gene expression data; and processing the current gene expression data and the current auxiliary information using at least part of the network layers in a prediction model trained according to the training method of the first aspect to obtain a correction value of the current gene expression data at the expected sequencing depth or an intermediate processing result of the correction value.

[0006] According to a third aspect of the present disclosure, a downstream task execution method is provided, comprising: acquiring input data, wherein the input data includes i) a correction value obtained according to the method of the second aspect; or ii) an intermediate processing result of the correction value in the method of the second aspect; or iii) a preprocessing result obtained by preprocessing the correction value obtained according to the method of the second aspect; or iv) a preprocessing result obtained by preprocessing the intermediate processing result of the correction value obtained according to the method of the second aspect; and using a downstream task algorithm to input data to obtain a downstream task result, wherein the downstream task includes a cell classification task, a perturbation prediction task or a drug response prediction task.

[0007] According to a fourth aspect of the present disclosure, a training device for a prediction model is provided, comprising: a first unit, for acquiring a plurality of samples, wherein each of the plurality of samples comprises first gene expression data, masked gene expression data and auxiliary information, the first gene expression data comprises counts of different genes in a single cell measured at a first sequencing depth, the masked gene expression data is obtained by downsampling the first gene expression data and masking the counts of some genes in the downsampled first gene expression data, the downsampled first gene expression data simulates the counts of different genes in a single cell measured at a second sequencing depth less than the first sequencing depth, the auxiliary information comprises a first total count, and the first total count is the sum of the counts of each gene in the first gene expression data; a second unit, for each of the plurality of samples: using the prediction model to be trained to process the masked gene expression data and the auxiliary information in the sample to obtain a predicted value of the counts of each gene at the first sequencing depth corresponding to the sample; determining a loss value corresponding to the sample according to the predicted value and the counts corresponding to some genes in the first gene expression data; and a third unit, for updating the prediction model to be trained according to the loss value corresponding to each of the plurality of samples.

[0008] According to a fifth aspect of the present disclosure, a correction device for gene expression data includes: a fourth unit, used to obtain current gene expression data and current auxiliary information, wherein the current gene expression data includes counts of different genes measured at an actual sequencing depth, and the current auxiliary information includes an expected first total count, the expected first total count is used to characterize the expected sequencing depth, and the expected first total count is greater than or equal to the sum of the counts of each gene in the current gene expression data; and a fifth unit, used to process the current gene expression data and the current auxiliary information using at least part of the network layer in a prediction model trained by the training device according to the fourth aspect, so as to obtain a correction value of the current gene expression data at the expected sequencing depth or an intermediate processing result of the correction value.

[0009] According to a sixth aspect of the present disclosure, a downstream task execution device includes: a sixth unit, used to obtain input data, wherein the input data includes i) a correction value obtained by the device according to the fifth aspect; or ii) an intermediate processing result of the correction value in the device according to the fifth aspect; or iii) a preprocessing result obtained by preprocessing the correction value obtained by the device according to the fifth aspect; or iv) a preprocessing result obtained by preprocessing the intermediate processing result of the correction value obtained by the device according to the fifth aspect; and a seventh unit, used to use the input data of a downstream task algorithm to obtain a downstream task result, wherein the downstream task includes a cell classification task, a perturbation prediction task or a drug response prediction task.

[0010] According to another aspect of the present disclosure, an electronic device is provided, including: a processor; and a memory, wherein the memory stores instructions executable by the processor, and when the instructions are executed by the processor, the processor executes any one of the above methods.

[0011] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing instructions is provided. When the instructions are executed by a processor, the processor executes any of the above methods.

[0012] According to another aspect of the present disclosure, a computer program product is provided, comprising: instructions, wherein when the instructions are executed by a processor, the processor is caused to execute any of the above methods.

[0013] These and other aspects of the disclosure will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0015] Figure 1A is a flowchart of a method for training a prediction model according to an embodiment of the present disclosure;

[0016] Figure 1B is a schematic diagram of a model structure of a prediction model according to an embodiment of the present disclosure;

[0017] Figure 2 is a flow chart of a method for correcting gene expression data according to an embodiment of the present disclosure;

[0018] Figure 3 is a flowchart of a downstream task execution method according to an embodiment of the present disclosure;

[0019] Figure 4 is a block diagram of a training device for a prediction model according to an embodiment of the present disclosure;

[0020] Figure 5 is a block diagram of a correction device for gene expression data according to an embodiment of the present disclosure;

[0021] Figure 6 is a block diagram of a downstream task execution device according to an embodiment of the present disclosure; and

[0022] Figure 7 is a block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure. DETAILED DESCRIPTION

[0023] In the present disclosure, unless otherwise specified, the use of the terms "first", "second", etc. to describe various elements is not intended to limit the positional relationship, timing relationship, or importance relationship of these elements, and such terms are only used to distinguish one element from another element. In some examples, the first element and the second element may refer to the same example of the element, and in some cases, based on the description of the context, they may also refer to different examples.

[0024] The terms used in the description of various examples described in this disclosure are only for the purpose of describing specific examples and are not intended to be limiting. Unless the context clearly indicates otherwise, if the number of elements is not specifically limited, the element can be one or more. As used herein, the term "plurality" means two or more, and the term "based on" should be interpreted as "based at least in part on". In addition, the terms "and / or" and "at least one of..." cover any one of the listed items and all possible combinations.

[0025] In the field of genomics, genes have expression levels in cells, and each cell has its true gene expression level. By sequencing cells, the number of reads aligned to each gene at a certain sequencing depth can be obtained. The number of reads aligned to a gene can be called the count of the gene, and the count can be understood as the gene expression observed at the sequencing depth. The relative relationship between the counts of each gene measured under different sequencing methods and sequencing depths may be different. It is understandable that the higher the sequencing depth, the higher the probability that the relative relationship between the observed counts of each gene is close to the relative relationship between the true gene expression levels of each gene in the cell. For example, when the sequencing depth is too low, the observed counts of some genes are 0, but when the sequencing depth is high enough, the observed counts are no longer 0. However, due to various factors, sometimes only a low sequencing depth is used in actual sequencing instead of the expected high sequencing depth. At the same time, due to the existence of technical noise, the accuracy of the relative relationship between the counts of each gene measured does not meet the requirements.

[0026] Therefore, we hope to be able to predict the relative relationship between the corrected gene counts through a prediction model based on the relative relationship between the gene counts observed at the actual sequencing depth, and use the relative relationship between the corrected gene counts as the relative relationship between the true gene expression levels of each gene in the cell for various downstream tasks.

[0027] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which the present disclosure belongs. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the relevant art and / or the context of this specification, and will not be interpreted in an idealized or overly formal sense, unless explicitly defined as such herein.

[0028] Figure 1A FIG. 1 is a flow chart of a method 100 for training a prediction model according to an embodiment of the present disclosure. Figure 1A As shown, the method 100 includes steps 110 to 130 .

[0029] In step 110, multiple samples are obtained. Each sample in the multiple samples includes first gene expression data, masked gene expression data and auxiliary information. The first gene expression data includes the counts of different genes in a single cell measured at a first sequencing depth. The masked gene expression data is obtained by downsampling the first gene expression data and masking the counts of some genes in the downsampled first gene expression data. The downsampled first gene expression data simulates the counts of different genes in a single cell measured at a second sequencing depth lower than the first sequencing depth. The downsampled first gene expression data is obtained by downsampling the first gene expression data. The auxiliary information includes a first total count. The first total count is the sum of the counts of each gene in the first gene expression data.

[0030] Herein, the first sequencing depth refers to the actual sequencing depth. The first gene expression data is the original gene expression data obtained by sequencing, that is, the counts of different genes in a single cell measured at the first sequencing depth. Table 1 is an example of the first gene expression data, and the corresponding first total count is the sum of the counts of each gene in Table 1. Exemplarily, the first gene expression data may include counts of N (about 20,000) genes, and the gene expression data of each sample are composed of the counts of the same genes.

[0031] Table 1

[0032] GeneID Gene1 Gene2 Gene3 Gene4 Gene5 … GeneN count 4 2 1 1 0 2

[0033] When the first gene expression data is obtained by transcript sequencing (eg, RNA-Seq technology), the first total count may also be referred to as a first total transcript number, that is, the sum of the transcript numbers of each gene in the first gene expression data.

[0034] Taking the first gene expression data as transcript sequencing results as an example, we will exemplify how the first gene expression data is collected. The scRNA-seq results of cells are stored in databases such as the Gene Expression Comprehensive Database, the Human Cell Atlas, and EMBL-EBI. scRNA-seq data were manually collected from these databases, and datasets with duplicate IDs were deleted. Most datasets provide raw count matrices. For datasets with standardized expression profiles, we convert them back to raw count form: we regard the smallest non-zero value in the original matrix as the raw count value 1, divide all remaining non-zero values ​​by the minimum value, and take the integer part. For datasets with TPM or FKPM expression profiles that cannot be converted back to raw counts, we keep them unchanged. In this way, we collected transcript sequencing results of more than 50 million single cells in multiple organs (such as heart, kidney, brain) and tissues (i.e., connective tissue, epithelial tissue, muscle tissue, and neural tissue).

[0035] There are usually 500 non-zero gene counts in the single-cell sequencing results. For example, the single-cell sequencing results whose non-zero gene counts exceed a non-zero count threshold (such as 200) can be screened out from the collected single-cell sequencing results as the first gene expression data to filter out cells with extremely low quality or damage.

[0036] When training the prediction model, the first gene expression data obtained by actual sequencing is used as the true value of the gene count at a high sequencing depth, the first gene expression data after downsampling is used to simulate the gene count at a low sequencing depth, the model is used to predict the gene count at a high sequencing depth based on the gene count at a low sequencing depth, and the loss is calculated based on the predicted value and the true value of the gene count at a high sequencing depth, and the model is updated with the loss value. In this way, the prediction model can learn to capture the relationship between the gene expressions of similar cells at different sequencing depths. In order to further improve the training effect, the MAE method is used for training, that is, the first gene expression data after downsampling is used as the gene count at a low sequencing depth, and partially masked, and the model is used to predict the gene count at a high sequencing depth based on the partially masked gene count at a low sequencing depth, and the loss is calculated based on the predicted gene count at a high sequencing depth and the corresponding part of the masked elements in the true value of the gene count at a high sequencing depth, and the model is updated with the loss value. Therefore, when constructing the training data, it is necessary to use the first gene expression data obtained by actual sequencing, downsample the first gene expression data, and mask the gene expression data obtained by masking the counts of some genes in the downsampled first gene expression data.

[0037] In order for the prediction model to correctly predict the gene count at a high sequencing depth based on the gene count at a low sequencing depth, the prediction model needs to understand what the low sequencing depth and high sequencing depth are. It is understandable that the higher the sequencing depth, the greater the sum of the counts of each gene in the sequencing result, so the sequencing depth can be characterized by the sum of the counts of each gene at that sequencing depth.

[0038] In one example, the auxiliary information includes the first total count, but does not necessarily include the sum of the gene counts at the low sequencing depth, that is, the sum of the counts of each gene in the downsampled first gene expression data. Because even if the auxiliary information does not include the sum of the counts of each gene in the downsampled first gene expression data, the model can guess the gene counts of the masked part based on the masked gene expression data in the sample, and then estimate the sum of the counts of each gene in the downsampled first gene expression data. Of course, the auxiliary information may also include the sum of the counts of each gene in the downsampled first gene expression data, so that the model can be explicitly informed of the low sequencing depth, thereby reducing the learning cost and accelerating model convergence.

[0039] The first sequencing depth refers to the higher sequencing depth used in actual sequencing, and the second sequencing depth refers to the lower sequencing depth simulated by the downsampled first gene expression data. The first sequencing depth and the second sequencing depth are used to indicate the sequencing depths corresponding to the first gene expression data and the mask expression data in the sample, respectively, and do not limit all samples to have the same first sequencing depth or the same second sequencing depth.

[0040] It is understandable that if the first gene expression data is measured at a sufficiently high sequencing depth, downsampling it can obtain a more credible gene count at a low sequencing depth. However, if the first gene expression data is originally measured at a low sequencing depth, the gene count at a low sequencing depth obtained by downsampling it is less credible. Therefore, in one example, the sum of the counts of each gene in the first gene expression data of each sample in the multiple samples in step 110 (that is, the sample whose first gene expression data is downsampled) is greater than a preset threshold, such as 1000. In one example, for the case where the sum of the counts of each gene in the first gene expression data is not greater than the preset threshold, only the first gene expression data can be masked to obtain masked gene expression data without downsampling. This type of sample includes the first gene expression data, the masked gene expression data, and auxiliary information. Unlike the sample in step 110, the masked gene expression data in this type of sample is obtained by masking the counts of some genes in the first gene expression data, without downsampling. Such samples can also be used to train a prediction model in a similar manner as the samples in step 110 . Through such samples, the model can learn to capture the relationship between genes in a single cell.

[0041] Since a single cell will be damaged after one sequencing, it is impossible to repeat sequencing of the same cell to obtain gene expression data at high and low sequencing depths, and thus it is impossible to construct gene expression data at low sequencing depth for each cell - gene expression data at high sequencing depth through actual sequencing methods. The only way is to simulate the gene expression data of the cell at low sequencing depth, i.e., the first gene expression data after downsampling, based on the gene expression data at high sequencing depth through algorithms, so as to construct the first gene expression data after downsampling (equivalent to the gene expression data at low sequencing depth) - first gene expression data (i.e., gene expression data at high sequencing depth) for each cell.

[0042] In one example, masked gene expression data can be obtained from the first gene expression data in the following manner: First, downsample the first gene expression data. Second, mask the counts of some genes in the downsampled first gene expression data to obtain masked gene expression data.

[0043] In one example, the masked gene expression data can also be obtained from the first gene expression data in the following manner: First, mask the counts of some genes in the first gene expression data. Second, downsample the masked first gene expression data to obtain the masked gene expression data.

[0044] The following explains how downsampling and masking are performed.

[0045] The downsampled first gene expression data is obtained by downsampling the first gene expression data using a statistical sampling algorithm, such as Poisson Distribution, Beta-Binomial Distribution, or Binomial Distribution.

[0046] In one example, a downsampling probability may be set, that is, the first gene expression data may be downsampled with a certain probability.

[0047] In one example, a sampling ratio can be set when downsampling is performed, and the sampling ratio represents the degree of downsampling of the first gene expression data. Setting a plurality of different sampling ratios can obtain first gene expression data that has been downsampled to different degrees from the same first gene expression data, thereby simulating gene counts at different low sequencing depths, thereby expanding the training data and achieving the augmentation of the training data. Exemplarily, the downsampled first gene expression data is obtained by downsampling the first gene expression data using a beta binomial distribution. The beta binomial distribution can automatically select a sampling ratio (sampling factor) based on the beta distribution.

[0048] For example, three downsampled first gene expression data with different downsampling degrees can be obtained from one first gene expression data.

[0049] The masking operation can be implemented by replacing the counts of the masked part of the genes in the first gene expression data with mask symbols. For example, the masked gene data can be [4, 1, M, M, 0, M, ..., M], wherein the gene counts of the masked part are replaced with "M", while the gene counts not masked retain the original values. In the obtained masked gene expression data, the gene ID corresponding to the masked gene count is known, and the gene count is unknown.

[0050] The number of masked gene counts in the masked gene expression data can be controlled by setting the masking ratio. The larger the masking ratio, the more gene counts are masked.

[0051] It should be noted that the specific covering ratio and covering method can be set according to actual needs and algorithm design, and the above is only an example.

[0052] In step 120, for each sample, the operations of step 121 and step 122 are performed. It can be understood that step 121 and step 122 can be sub-steps of step 120.

[0053] In step 121, the masked gene expression data and auxiliary information in the sample are processed using the prediction model to be trained to obtain a predicted value of each gene count at a first sequencing depth corresponding to the sample.

[0054] In one example, the architecture of the prediction model to be trained can adopt an encoder-decoder network, a decoder network, or a multi-layer perceptron. The prediction model can include more than 1 million parameters, such as 3 million, 10 million, or 100 million parameters. The model parameters can be increased by increasing the number of layers and / or the number of hidden vectors per layer.

[0055] Exemplarily, when the prediction model to be trained is used to process the masked gene expression data and auxiliary information in the sample, the masked gene expression data and the auxiliary information can both be input from the first network layer of the prediction model, or the masked gene expression data can be input from the first network layer of the prediction model, and the auxiliary information can be input from the later network layer of the prediction model; the masked gene expression data and the auxiliary information themselves can be input into the prediction model, or the masked gene expression data and the auxiliary information can be converted into corresponding embedding vectors and then input into the prediction model; the counts corresponding to all genes in the masked gene expression data can be input into the first network layer of the prediction model, or the counts corresponding to some genes (for example, the counts corresponding to genes whose counts are not 0 and are not masked).

[0056] The masked gene expression data, the feature tensor corresponding to the masked gene expression data, the concatenation result of the masked gene expression data and the auxiliary information, and the concatenation result of the feature tensor corresponding to the masked gene expression data and the feature tensor corresponding to the auxiliary information as the input feature tensor can be interpreted in a broad sense, not only referring to the first network layer of the input prediction model, but also referring to any network layer of the input prediction model. For example, part of the tensor in the input feature tensor is used as the input of the first network layer of the prediction model, and part of the tensor is used as the input of the intermediate network layer of the prediction model. In addition, the input feature tensor may include the part corresponding to all elements of the masked gene expression data, or may only include the part corresponding to some elements of the masked gene expression data (such as non-zero and uncovered parts), and also include the part corresponding to elements outside the masked gene expression data (such as filled elements, which will be explained later).

[0057] It should be noted that the size of the object to be processed of the network layer of the input prediction model is certain, for example, 2000*768. When the counts corresponding to the genes with non-zero counts and not covered in the masked gene expression data are input into the network layer of the prediction model, the different numbers of non-zero gene counts and the different proportions of covered elements included in different masked gene expression data may lead to different sizes of the objects to be processed of the network layer of the input prediction model. In this case, they can be padded to the same size to ensure that the sizes of the objects to be processed of the network layer of the input prediction model are the same. For example, the number of non-zero and uncovered gene counts included in the masked gene expression data in each sample fluctuates between 300-1000, and the size of the input feature tensor can be padded to 1000. For example, the size of the input feature tensor can be padded to 10% of the total number of genes.

[0058] The following is an example of how to convert masked gene expression data into its corresponding embedding vector.

[0059] Exemplarily, an embedding vector can be created for each element in the masked gene expression data to obtain the embedding vector corresponding to the masked gene data. The embedding vector of each element can be obtained by adding the gene embedding vector representing the gene ID and the count embedding vector representing the gene count. There are three types of elements for masked gene expression data: elements with non-zero counts and not masked, elements with counts of 0, and masked elements. They all have gene IDs and can determine the gene embedding vector representing the gene ID. For elements with a count of 0, the count embedding vector representing the gene count can be randomly generated. For masked elements, the count embedding vector representing the gene count can be a specified embedding vector. For elements with a count of non-0, the count embedding vector representing the gene count can be determined by a lookup table. For example, the lookup table can record the correspondence between the count value range and the count embedding vector, and count values ​​in different ranges correspond to different count embedding vectors.

[0060] Exemplarily, the count embedding vector can be determined by a module for determining the count embedding vector according to the count, and the module includes learnable parameters. In particular, a lookup table is randomly initialized, which includes 100 1*768 vectors. For a count x, first pass through a linear layer and a leaky ReLU layer to obtain an intermediate vector v1=leaky ReLU(x*w1), where w1 is a learnable parameter with a size of 1*100. Afterwards, the intermediate vector v1 is processed with another linear layer and a scaling factor to obtain an intermediate vector v2. v2=v1*w2+α*v2. The size of v2 is 100*100, and α is a learnable parameter. Afterwards, v2 is normalized with SoftMax to obtain a weight vector v3, and the size of v3 is 1*100. With v3 as the weight of each of the 100 vectors, the 100 vectors in the lookup table are weighted and summed to obtain the count vector Ex. Among them, the learnable parameters in the module for determining the count embedding vector according to the count can be trained together with the prediction model.

[0061] As mentioned above, the input feature tensor may include parts corresponding to padding elements. In this case, a padding vector can be assigned to the padding element as its corresponding embedding vector, and no longer distinguish between gene embedding vectors and count embedding vectors.

[0062] Exemplarily, for the total count in the auxiliary information, its corresponding embedding vector can also be determined in a manner similar to the elements in the masked gene expression data. A symbol similar to the gene ID is assigned to the total count in the auxiliary information, and the gene embedding vector is determined based on the symbol. The numerical value of the total count is regarded as the count, and the count embedding vector is determined based on the total count, and then the two are added bit by bit.

[0063] Exemplarily, the masked gene expression data and the auxiliary information may both be input from the first network layer of the prediction model. In this case, step 121 includes:

[0064] The masked gene expression data in the sample is concatenated with the auxiliary information as the input feature tensor corresponding to the sample; or the feature tensor corresponding to the masked gene expression data in the sample is concatenated with the feature tensor corresponding to the auxiliary information as the input feature tensor corresponding to the sample; and the input feature tensor is processed using the prediction model to be trained.

[0065] Exemplarily, the masked gene expression data can be input from the first network layer of the prediction model, and the auxiliary information can be input from the network layer behind the prediction model. The prediction model to be trained includes an input layer, an output layer, and multiple intermediate layers between the input layer and the output layer. At this time, step 121 includes: using the masked gene expression data in the sample as an input feature tensor or the feature tensor corresponding to the masked gene expression data in the sample as an input feature tensor, inputting the input layer and the first predetermined layer in the multiple intermediate layers to obtain an intermediate feature tensor, and the number of the first predetermined layers is greater than or equal to 0; splicing the auxiliary information with the intermediate feature tensor to obtain the spliced ​​intermediate feature tensor; and inputting the spliced ​​intermediate feature tensor into the second predetermined layer in the multiple intermediate layers and the output layer, and the second predetermined layer is different from the first predetermined layer. For example, when the prediction model includes a decoder network, the auxiliary information is input before the last several layers of the decoder network; for another example, when the prediction model includes an encoder network and a decoder network, the auxiliary information can be input before a certain layer of the encoder network or before the last several layers of the decoder network. In this way, the model can use the auxiliary information to extract more meaningful features or make more accurate predictions in the process of processing input data.

[0066] It is understandable that when the architecture of the prediction model is a decoder or encoder-decoder, the prediction model may also include a multilayer perceptron after the decoder, which is used to project the decoding tensor output by the decoder into the predicted value of each gene count at the first sequencing depth.

[0067] It is understandable that the input feature tensors of multiple samples can be superimposed and simultaneously input into the first network layer of the prediction model. The superimposed input feature tensor can be as shown in Table 2.

[0068] Table 2

[0069]

[0070] In step 122, a loss value corresponding to the sample is determined according to the predicted value and the counts corresponding to the partial genes in the first gene expression data.

[0071] In one example, the loss value is determined by using the counts corresponding to the masked portion of genes in the predicted value and the true value.

[0072] In step 130, the prediction model to be trained is updated according to the loss value corresponding to each sample.

[0073] Exemplarily, the prediction model to be trained may be updated with the loss of a single sample, or the prediction model to be trained may be updated with the sum of the losses of multiple samples.

[0074] Method 100 downsamples the actual sequencing results at a high sequencing depth, simulates the sequencing results at a low sequencing depth, uses the sequencing results at the high sequencing depth and the sequencing results at the low sequencing depth as samples to train the prediction model, and explicitly introduces auxiliary information that characterizes the expected sequencing depth, so that the prediction model learns to capture the relationship between gene counts of similar cells at different sequencing depths and the relationship between gene counts in a single cell, thereby minimizing technical noise interference in sequencing, and increasing the sequencing depth by computation, thereby improving the accuracy of the output results of downstream tasks.

[0075] In one example, each sample in the plurality of samples includes normalized first gene expression data, normalized masked gene expression data, and auxiliary information. The normalized first gene expression data is obtained by normalizing the first gene expression data, and the normalized masked gene expression data is obtained by downsampling the first gene expression data, normalizing the downsampled first gene expression data, and masking the counts of some genes in the normalized result. In addition to the first total count T, the auxiliary information also includes a second total count S. The second total count is the sum of the counts of each gene in the downsampled first gene expression data.

[0076] Compared with the absolute counts of genes, the relative relationship between the counts of different genes is more important for downstream tasks. In order to eliminate the differences in gene counts caused by sequencing depth and make the sequencing results of different samples and different sequencing depths comparable, the sequencing results need to be standardized and normalized to obtain the samples used for training. In other words, at this time, the normalized first gene expression data and the normalized masked gene expression data can be used to replace the first gene expression data and the masked gene expression data in step 110, step 120, step 121, and step 122.

[0077] The sequencing depth is represented by the total counts. In order to eliminate the influence of sequencing depth, the total counts in the gene expression data can be used for normalization.

[0078] In one example, the first gene expression data is divided by the sum of the counts of each gene in the first gene expression data for normalization.

[0079] In one example, a normalization method such as TPM or RPKM may be used to normalize the first gene expression data or the downsampled first gene expression data.

[0080] In one example, the log(TPM+1) normalization method can also be used to normalize the first gene expression data and the downsampled first gene expression data. For example, the first gene expression data is a vector X1, and the normalized first gene expression data is log[X1 / sum(X1)*10000+1]. The downsampled first gene expression data is a vector X2, and the downsampled and normalized first gene expression data is log[X2 / sum(X2)*10000+1].

[0081] The log(TPM+1) normalization method performs a logarithmic transformation on the basis of TPM. Logarithmic transformation can reduce the skewness in the original data and improve the comparability between variables. In this method, a constant 1 is added to the TPM of each first gene expression data, and then the logarithm is taken. The final result is a real number representing the TPM value after logarithmic transformation. In short, the log(TPM+1) normalization method is a variant method of the standard TPM normalization method that performs a logarithmic transformation, which can eliminate skewness and improve the comparability between variables.

[0082] The downsampled first gene expression data simulates the counts of different genes in a single cell measured at a second sequencing depth lower than the first sequencing depth, and the second total count can be used as a representation of the second sequencing depth. Although the first sequencing depth can be obtained through actual sequencing and is known, the second sequencing depth is unknown. In order to characterize the first sequencing depth and the second sequencing depth with the same measure, the first total count and the second total count are used to characterize the first sequencing depth and the second sequencing depth.

[0083] Exemplarily, the first total count T and the second total count S can both determine their corresponding embedding vectors in the manner described above, and the sign of the first total count is different from the sign of the second total count and is also different from other gene IDs.

[0084] In the example, since the sample only includes the normalized masked gene expression data, but does not include the downsampled first gene expression data before normalization, the model cannot estimate the sum of the counts of each gene in the downsampled but not normalized first gene expression data, that is, it cannot estimate how low the lower sequencing depth corresponding to the normalized masked gene expression data is. Therefore, the second total count must be included in the auxiliary information. In this way, the prediction model can explicitly know how low the lower sequencing depth corresponding to the normalized masked gene expression data is and how high the higher sequencing depth corresponding to the first gene expression data is based on the second total count, and then be able to predict the gene counts at a higher sequencing depth based on the normalized masked gene expression data.

[0085] It is understandable that, since the normalized first gene expression data is used as the true value, the predicted value of each gene count at the first sequencing depth output by the model is also normalized. The predicted value can be restored to the predicted value of the gene count that has not been normalized by multiplying the predicted value by the first total count, thereby remapping the normalized predicted value to the original count space.

[0086] Obtaining normalized masked gene expression data from the first gene expression data requires the steps of normalization, downsampling, and masking, and the order of the steps can be combined in various ways.

[0087] In one example, the normalized masked gene expression data can be obtained from the first gene expression data in the following manner: Step 1, downsampling the first gene expression data. Step 2, normalizing the downsampled first gene expression data. Step 3, masking the counts of some genes in the result of Step 2 to obtain the masked gene expression data.

[0088] In one example, the normalized masked gene expression data can also be obtained from the first gene expression data in the following manner: In the first step, the first gene expression data is normalized. In the second step, the normalized first gene expression data is downsampled. In the third step, the result of the second step is masked to obtain the normalized masked gene expression data.

[0089] When each of the multiple samples includes normalized first gene expression data, normalized mask gene expression data, and auxiliary information, the implementation process of steps 121 and 122 is similar to the above steps and will not be repeated herein.

[0090] In one example, the prediction model to be trained includes an encoder network and a decoder network. Step 121 can be implemented by the following steps.

[0091] Exemplarily, before step 1211, the masked gene expression data and the auxiliary information can be input into a module for determining count embedding vectors according to counts using the method described above to obtain an input feature tensor. For example, the size of the masked gene expression data is N*1, and the size of the input feature tensor is N*768.

[0092] Step 1211, using the encoder network to encode the first tensor corresponding to the unmasked and non-zero counts in the input feature tensor to obtain an encoded tensor. Exemplarily, the size of the first tensor can be 2000*768, and the size of the encoded tensor can be 2000*768. The 1st, 2nd, 4th, 10th, 30th, etc. elements of the 2000 elements are non-zero and unmasked elements, and the 1561st to 2000th elements are filled elements.

[0093] Step 1212: Replace the first tensor in the input feature tensor with the encoding tensor to obtain the encoded input feature tensor. Replace the 1st, 2nd, 4th, 10th, 30th, etc. elements in N*768 with the corresponding tensors in the encoding tensor.

[0094] Step 1213: decode the encoded input feature tensor using a decoder network to obtain a decoded tensor. The size of the decoded tensor may be N*768.

[0095] Exemplarily, the prediction model also includes a multilayer perceptron. After step 1213, the multilayer perceptron in the prediction model is used to process the decoded tensor to obtain the predicted value of each gene count at the first sequencing depth corresponding to the sample. The size of the predicted value is N*1. It is understandable that the multilayer perceptron is trained together with other parts in the prediction model.

[0096] The process described in this embodiment is as follows Figure 1B shown.

[0097] The encoder part focuses on the embedding vectors corresponding to the non-zero counts in the input feature tensor, while the decoder part receives the embedding vectors of all genes (i.e., the embeddings processed by the encoder and the embeddings of other genes) and integrates the information of all positions. After the zero-value positions and masked positions are filtered before being fed to the encoder, the input sequence length of the encoder input is approximately 10% of the full gene length. This design greatly reduces the required computing resources, enabling the encoder to use a series of ordinary Transformer blocks to capture gene dependencies, greatly improving training efficiency and training effects. Since the encoder module only processes non-zero, unmasked counts, the model can effectively focus on the most informative non-zero expressed genes, while allowing zero-value genes to participate in model training in the decoder stage.

[0098] In one example, the encoder includes M layers of encoding units, and the decoder includes N layers of decoding units. The value of M is greater than the value of N.

[0099] Encoders often require more layers to extract high-order features of data to better represent the original data. The decoder can use fewer layers to complete the decoding task because it only needs to restore the low-dimensional feature representation to the original data, and does not need to extract and abstract features like the encoder. Therefore, when setting the number of layers of the encoder and decoder, you can choose a deeper encoder and a shallower decoder, which can reduce the number of model parameters while retaining the original data information, and improve the training efficiency and generalization ability of the model.

[0100] Furthermore, each layer of the encoding unit of the encoder includes a multi-head attention unit and a forward propagation unit. Each layer of the decoding unit of the decoder includes a forward propagation unit and also includes a linear attention unit or a sparse attention unit.

[0101] Using multi-head attention units and forward propagation units as encoding units in each layer of the encoder can help the model better capture the dependencies and semantic information between different genes, further improving the performance of the model. First, gene expression data is usually high-dimensional and sparse, with high noise and complexity, so the encoder's multi-head attention unit can help the model better mine the associations and interactions between genes.

[0102] In the decoder, using the forward propagation unit and the linear attention unit or the sparse attention unit can help the model better predict the first gene expression data gene by gene. The linear attention unit can help the model better control the distribution of attention weights and improve the performance and stability of the model. The sparse attention unit can further reduce the amount of computation of the model and improve the efficiency of the model.

[0103] In general, the encoder-decoder model using multi-head attention units, forward propagation units, and attention mechanisms can achieve better performance and results in predicting gene expression data. This model can automatically learn complex patterns and associations in the data and accurately predict and analyze the first gene expression data.

[0104] In one example, in batch training, a batch of training examples is fed into the model, the batch loss is calculated, and then the parameters are updated through backpropagation. The updates can be done one by one or simultaneously.

[0105] In one example, the preset training end condition may include any of the following: the loss value is less than the loss value threshold, the number of training times reaches a preset number of times, and the training time reaches a preset duration.

[0106] Figure 2 is a flow chart of a method for correcting gene expression data according to an embodiment of the present disclosure. Figure 2 As shown, the method 200 includes step 210 and step 220 .

[0107] In step 210, current gene expression data and current auxiliary information are obtained, wherein the current gene expression data includes counts of different genes measured at an actual sequencing depth, and the current auxiliary information includes an expected first total count, which is used to characterize the expected sequencing depth, and the expected first total count is greater than or equal to the sum of the counts of each gene in the current gene expression data.

[0108] It is understandable that when the expected first total count is greater than the sum of the counts of each gene in the current gene expression data, it is expected that the prediction model predicts the expected gene count at a higher sequencing depth based on the gene count actually measured at a lower sequencing depth. When the expected first total count is equal to the sum of the counts of each gene in the current gene expression data, it is expected that the prediction model predicts the corrected gene count under the condition that the sequencing depth remains substantially unchanged based on the gene count actually measured.

[0109] Since the space for increasing sequencing depth is limited, and in order to ensure the data accuracy of the predicted value of each gene count in the target single cell at the expected sequencing depth, it is expected that the first total count cannot be much larger than the sum of the counts of each gene in the current gene expression data.

[0110] The expected sequencing depth can be set according to the sequencing method. For example, if the sequencing method is 10X Genomics Chromium sequencing technology (10X Genomics Chromium sequencing technology, referred to as "10X sequencing technology"), the sum of the counts of each gene in the current gene expression data is about 1000, and the expected sequencing depth can be set to 10000.

[0111] It is understandable that the current gene expression data may be single-cell gene expression data or gene expression data obtained by sequencing a large number of cells (bulk).

[0112] In step 220 , the current gene expression data and the current auxiliary information are processed using at least part of the network layers in the prediction model to obtain a correction value or an intermediate processing result of the correction value of the current gene expression data at the desired sequencing depth. Here, the prediction model can be trained according to method 100 .

[0113] It is understandable that in the inference stage, all network layers in the prediction model can be used to process the current gene expression data and the current auxiliary information, and what is obtained at this time is the correction value of the current gene expression data at the expected sequencing depth; or only part of the network layers in the prediction model (such as the first M layers) can be used to process the current gene expression data and the current auxiliary information, and what is obtained at this time is the intermediate processing result of the correction value of the current gene expression data at the expected sequencing depth. It depends on the needs of the downstream tasks.

[0114] Exemplarily, the correction value is Figure 1B The output of the multilayer perceptron in the middle, the intermediate processing result of the correction value refers to the intermediate processing result obtained to obtain the correction value, and does not mean that the final correction value is obtained. The intermediate processing result can be Figure 1B The encoding vector output by the encoder, the decoding vector output by the decoder, etc.

[0115] For example, continue to refer to Figure 1BThe prediction model includes an encoder and a decoder. The encoding vector output by the encoder can be used to obtain the features representing the cells, and the decoding vector output by the decoder can be used to obtain the features representing the genes. When the downstream task needs to use cell features, in the inference stage, only the encoder can be used, and the features representing the cells obtained according to the encoding vector can be used for the downstream task; when the downstream task needs to use gene features, in the inference stage, the encoder and decoder can be used, and the features representing the genes obtained according to the decoding vector can be used for the downstream task.

[0116] Exemplarily, the encoding vector can be passed through a pooling layer (e.g., a maximum pooling layer) to obtain a feature representing a cell. The feature representing a cell can be 2000*1. The decoding vector can be passed through a multi-layer perceptron to obtain a feature representing a gene, and the feature representing a gene can be N*1.

[0117] In one example, the current auxiliary information also includes a current second total count, which is the sum of the counts of each gene in the current gene expression data.

[0118] Further, step 210 can be implemented by the following method: obtaining normalized current gene expression data and current auxiliary information. The normalized current gene expression data is obtained by normalizing the current gene expression data. The normalization operation can be specifically described in step method 100, which will not be repeated here.

[0119] Furthermore, step 220 can be implemented in the following manner: using a prediction model to process the normalized current gene expression data and the current auxiliary information to obtain a predicted value of the normalized value of each gene count of the target single cell at the desired sequencing depth.

[0120] Method 200 can improve the accuracy of the relative relationship between gene counts through a prediction model.

[0121] Exemplarily, the correction value predicted in step 200 or the intermediate processing result of the correction value may be used in a variety of downstream tasks to improve the accuracy of the downstream tasks.

[0122] Exemplarily, the cell features or gene features obtained according to method 200 may replace the cell features or gene features originally used in the downstream task.

[0123] Figure 3 is a flow chart of a downstream task execution method according to an embodiment of the present disclosure. Figure 3 As shown, the method 300 includes step 310 and step 320 .

[0124] In step 310, input data is obtained, wherein the input data includes i) a correction value obtained according to method 200; or ii) an intermediate processing result of the correction value obtained according to method 200; or iii) a preprocessing result obtained by preprocessing the correction value obtained by method 200; or iv) a preprocessing result obtained by preprocessing the intermediate processing result of the correction value obtained by method 200.

[0125] Exemplarily, preprocessing may be pooling, processing using a linear layer, processing using a specific model, and the like.

[0126] Exemplarily, downstream tasks include cell classification tasks (further subdivided into cell classification tasks, cell clustering tasks), perturbation prediction tasks or drug response prediction tasks. Exemplarily, features characterizing cells are used for cell classification tasks, cell clustering tasks, and drug response prediction tasks, and features characterizing genes are used for perturbation prediction tasks.

[0127] Exemplarily, when the downstream task is a cell classification task, the features characterizing the cells are used in the cell classification model to obtain the cell classification result. In the example, the classification model can be, for example, a support vector machine (SVM), a multilayer perceptron (MLP), or a decision tree.

[0128] For example, when the downstream task is a cell clustering task, a clustering algorithm can be used to cluster the features representing the cells to obtain a clustering result. The clustering algorithm can be K-means clustering or the like.

[0129] The cell classification results may be, for example, classification results of cancer cells and normal cells, clustering results of different types of immune cells, or classification results of cell types in different organs.

[0130] When the downstream task is a drug response prediction task, the graph neural network can be used to extract drug features, and the prediction model can be used to determine the features that characterize the cells. The drug features and the features that characterize the cells are input into the drug response prediction model to obtain the prediction result of whether the cells are sensitive to the drug.

[0131] When the downstream task is a perturbation prediction task, the features characterizing the genes determined by the prediction model can be used as the gene features corresponding to the gene nodes of the gene co-expression graph neural network. The perturbation features are applied to the gene co-expression graph neural network to predict the expression of the perturbed genes.

[0132] Exemplarily, models other than the prediction model used in downstream tasks can be trained together with the prediction model.

[0133] Figure 4 4 is a block diagram of a training device 400 for a prediction model according to an embodiment of the present disclosure. Figure 4 As shown, the prediction model training device 400 includes a first unit 410 , a second unit 420 and a third unit 430 .

[0134] The first unit 410 is used to obtain multiple samples, wherein each sample in the multiple samples includes first gene expression data, masked gene expression data and auxiliary information, the first gene expression data includes counts of different genes in a single cell measured at a first sequencing depth, the masked gene expression data is obtained by downsampling the first gene expression data and masking the counts of some genes in the downsampled first gene expression data, the downsampled first gene expression data simulates the counts of different genes in a single cell measured at a second sequencing depth less than the first sequencing depth, and the auxiliary information includes a first total count, which is the sum of the counts of each gene in the first gene expression data.

[0135] The second unit 420 is used for: for each sample: using the prediction model to be trained to process the masked gene expression data and auxiliary information in the sample to obtain the predicted value of each gene count at the first sequencing depth corresponding to the sample; and determining the loss value corresponding to the sample based on the predicted value and the counts corresponding to some genes in the first gene expression data.

[0136] The third unit 430 is used to update the prediction model to be trained according to the loss value corresponding to each sample.

[0137] It should be understood that Figure 4 The various units of the device 400 shown in FIG. 4 can be the same as those in the above reference Figure 1A The steps in the method 100 described above correspond to each other. Therefore, the operations, features and advantages described above for the method 100 are also applicable to the device 400 and the units included therein. For the sake of brevity, some operations, features and advantages are not repeated here.

[0138] Figure 5 is a block diagram of an apparatus 500 for predicting gene expression data according to an embodiment of the present disclosure. Figure 5 As shown, the apparatus 500 for predicting gene expression data includes a fourth unit 510 and a fifth unit 520 .

[0139] The fourth unit 510 is used to obtain current gene expression data and current auxiliary information, wherein the current gene expression data includes the counts of different genes in the target single cell measured at the actual sequencing depth, and the current auxiliary information includes the expected first total count, which is the sum of the counts of each gene at the expected sequencing depth, and the expected sequencing depth is greater than the actual sequencing depth.

[0140] The fifth unit 520 is used to utilize Figure 4 The prediction model trained by the training device processes the current gene expression data and the current auxiliary information to obtain the predicted value of each gene count of the target single cell at the expected sequencing depth.

[0141] It should be understood that Figure 5 The various units of the device 500 shown in FIG. 5 can be the same as those in the above reference Figure 2 The steps in the method 200 described above correspond to each other. Therefore, the operations, features and advantages described above for the method 200 are also applicable to the device 500 and the units included therein. For the sake of brevity, some operations, features and advantages are not repeated here.

[0142] Figure 6 FIG. 6 is a block diagram of a cell classification device 600 according to an embodiment of the present disclosure. Figure 6 As shown, the cell sorting device 600 includes a sixth unit 610 and a seventh unit 620 .

[0143] The sixth unit 610 is used to obtain input data, wherein the input data includes i) using a cell feature extraction model to extract Figure 5 ii) extracting the embedding vector from the predicted value in the device; or ii) according to Figure 5 iii) the intermediate processing results of the prediction model according to Figure 5 In the device, the intermediate processing results of the prediction model are preprocessed to obtain the processed intermediate processing results.

[0144] The seventh unit 620 is used to process the input data using a classification model or a clustering algorithm to obtain a cell classification result.

[0145] It should be understood that Figure 6 The various units of the device 600 shown in FIG. 6 can be the same as those described above with reference to FIG. Figure 3 The steps in the method 300 described above correspond to each other. Therefore, the operations, features and advantages described above for the method 300 are also applicable to the apparatus 600 and the units included therein. For the sake of brevity, some operations, features and advantages are not described in detail here.

[0146] Although specific functions are discussed above with reference to specific units, it should be noted that the functions of the various units discussed herein can be divided into multiple units, and / or at least some functions of multiple units can be combined into a single unit. The specific unit discussed herein performs an action including the specific unit itself performing the action, or alternatively the specific unit calls or otherwise accesses another component or unit that performs the action (or performs the action in conjunction with the specific unit). Therefore, the specific unit that performs an action can include the specific unit itself that performs the action and / or another unit that the specific unit calls or otherwise accesses and performs the action.

[0147] It should also be understood that various techniques may be described herein in the general context of software hardware elements or program units. Figure 4 , Figure 5 as well as Figure 6 The various units described may be implemented in hardware or in hardware in combination with software and / or firmware. For example, these units may be implemented as computer program code / instructions configured to be executed in one or more processors and stored in a computer-readable storage medium. Alternatively, these units may be implemented as hardware logic / circuits.

[0148] According to one aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and at least one memory communicatively connected to the at least one processor, wherein the at least one memory stores instructions, and when the instructions are executed by the at least one processor, the at least one processor executes the above method.

[0149] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing instructions is provided. When the instructions are executed by at least one processor of a computer, the computer executes the above method.

[0150] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, and the computer program implements the above method when executed by a processor.

[0151] Figure 7 is a block diagram of an exemplary electronic device that can be used to implement the embodiments of the present disclosure.

[0152] The electronic device 700 can be a variety of different types of devices. Examples of the electronic device 700 include, but are not limited to, a desktop computer, a server computer, a laptop or netbook computer, a mobile device (e.g., a tablet computer, a cellular or other wireless phone (e.g., a smart phone), a notepad computer, a mobile station), a wearable device (e.g., glasses, a watch), an entertainment device (e.g., an entertainment appliance, a set-top box communicatively coupled to a display device, a game console), a television or other display device, a car computer, and the like.

[0153] The electronic device 700 may include at least one processor 702, memory 704, communication interface(s) 706, a display device 708, other input / output (I / O) devices 710, and one or more mass storage devices 712 that can communicate with each other, such as via a system bus 714 or other appropriate connection.

[0154] The processor 702 may be a single processing unit or multiple processing units, all of which may include a single or multiple computing units or multiple cores. The processor 702 may be implemented as one or more microprocessors, microcomputers, microcontrollers, digital signal processors, central processing units, state machines, logic circuits, and / or any device that manipulates signals based on operating instructions. Among other capabilities, the processor 702 may be configured to obtain and execute computer-readable instructions stored in the memory 704, mass storage device 712, or other computer-readable media, such as program codes of an operating system 716, program codes of an application program 718, program codes of other programs 720, and the like.

[0155] The memory 704 and the mass storage device 712 are examples of computer-readable storage media for storing instructions that are executed by the processor 702 to implement the various functions described above. For example, the memory 704 may generally include both volatile memory and non-volatile memory (e.g., RAM, ROM, etc.). In addition, the mass storage device 712 may generally include a hard drive, a solid-state drive, a removable medium, including external and removable drives, a memory card, a flash memory, a floppy disk, an optical disk (e.g., a CD, a DVD), a storage array, a network attached storage, a storage area network, etc. The memory 704 and the mass storage device 712 may all be collectively referred to herein as memory or computer-readable storage media, and may be a non-transitory medium capable of storing computer-readable, processor-executable program instructions as computer program code, which may be executed by the processor 702 as a specific machine configured to implement the operations and functions described in the examples herein.

[0156] A number of programs may be stored on the mass storage device 712. These programs include an operating system 716, one or more application programs 718, other programs 720, and program data 722, and they may be loaded into the memory 704 for execution. Examples of such applications or program modules may include, for example, computer program logic (e.g., computer program code or instructions) for implementing the following components / functions: method 100 (including any suitable steps of method 100), method 200 (including any suitable steps of method 200), method 300 (including any suitable steps of method 300), and / or additional embodiments described herein.

[0157] Although in Figure 7 704 of the electronic device 700, but modules 716, 718, 720, and 722, or portions thereof, may be implemented using any form of computer-readable media accessible by the electronic device 700. As used herein, "computer-readable media" includes at least two types of computer-readable media, namely, computer-readable storage media and communication media.

[0158] Computer-readable storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information, such as computer-readable instructions, data structures, program modules or other data. Computer-readable storage media include but are not limited to RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD), or other optical storage device, magnetic cassette, magnetic tape, magnetic disk storage device or other magnetic storage device, or any other non-transmission medium that can be used to store information for access by electronic devices. In contrast, communication media can embody computer-readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transmission mechanism. Computer-readable storage media defined herein do not include communication media.

[0159] One or more communication interfaces 706 are used to exchange data with other devices, such as through a network, direct connection, etc. Such communication interfaces can be one or more of the following: any type of network interface (e.g., a network interface card (NIC)), a wired or wireless (such as IEEE 802.11 wireless LAN (WLAN)) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth™ interface, a Near Field Communication (NFC) interface, etc. The communication interface 706 can facilitate communication within a variety of network and protocol types, including wired networks (e.g., LAN, cable, etc.) and wireless networks (e.g., WLAN, cellular, satellite, etc.), the Internet, etc. The communication interface 706 can also provide communication with external storage devices (not shown) such as storage arrays, network attached storage, storage area networks, etc.

[0160] In some examples, a display device 708 such as a monitor may be included for displaying information and images to the user. Other I / O devices 710 may be devices that receive various inputs from the user and provide various outputs to the user, and may include a touch input device, a gesture input device, a camera, a keyboard, a remote control, a mouse, a printer, an audio input / output device, and the like.

[0161] The technology described herein can be supported by these various configurations of the electronic device 700, and is not limited to the specific examples of the technology described herein. For example, the function can also be implemented in whole or in part on the "cloud" by using a distributed system. The cloud includes and / or represents a platform for resources. The platform abstracts the underlying functions of the hardware (e.g., server) and software resources of the cloud. Resources can include applications and / or data that can be used when performing computing processing on a server away from the electronic device 700. Resources can also include services provided through the Internet and / or through a subscriber network such as a cellular or Wi-Fi network. The platform can abstract resources and functions to connect the electronic device 700 to other electronic devices. Therefore, the implementation of the functions described herein can be distributed throughout the cloud. For example, the functions can be implemented partially on the electronic device 700 and partially through a platform that abstracts the functions of the cloud.

[0162] Although the present disclosure has been illustrated and described in detail in the drawings and the foregoing description, such illustration and description should be considered illustrative and exemplary rather than restrictive; the present disclosure is not limited to the disclosed embodiments. By studying the drawings, the disclosure and the appended claims, those skilled in the art will be able to understand and implement variations to the disclosed embodiments when practicing the claimed subject matter. In the claims, the word "comprising" does not exclude other elements or steps that are not listed, the indefinite article "a" or "an" does not exclude a plurality, and the term "plurality" means two or more. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Claims

1. A training method for a prediction model, include: Acquire multiple samples, wherein each sample in the multiple samples includes first gene expression data, masked gene expression data, and auxiliary information, the first gene expression data includes counts of different genes in a single cell measured at a first sequencing depth, the masked gene expression data is obtained by downsampling the first gene expression data and masking counts of some genes in the downsampled first gene expression data, the downsampled first gene expression data simulates counts of different genes in a single cell measured at a second sequencing depth lower than the first sequencing depth, and the auxiliary information includes a first total count, which is the sum of counts of each gene in the first gene expression data; For each sample in the plurality of samples: Using the prediction model to be trained to process the masked gene expression data and auxiliary information in the sample to obtain a predicted value of each gene count at a first sequencing depth corresponding to the sample; Determining a loss value corresponding to the sample according to the predicted value and the counts corresponding to the part of genes in the first gene expression data; and The prediction model to be trained is updated according to the loss value corresponding to each sample in the multiple samples.

2. The method according to claim 1, in, Each sample of the multiple samples includes normalized first gene expression data, normalized masked gene expression data, and auxiliary information, wherein the normalized first gene expression data is obtained by normalizing the first gene expression data, and the normalized masked gene expression data is obtained by downsampling the first gene expression data, normalizing the downsampled first gene expression data, and masking the counts of some genes in the normalized result; the auxiliary information also includes a second total count, which is the sum of the counts of each gene in the downsampled first gene expression data; The method of processing the masked gene expression data and auxiliary information in the sample using the prediction model to be trained to obtain the predicted value of each gene count at the first sequencing depth corresponding to the sample includes: The normalized masked gene expression data in the sample and the auxiliary information are processed using the prediction model to be trained to obtain a predicted value of the normalized value of each gene count at the first sequencing depth corresponding to the sample, and Wherein, determining the loss value corresponding to the sample according to the predicted value and the counts corresponding to the part of genes in the first gene expression data includes: The loss value corresponding to the sample is determined according to the predicted value and the counts corresponding to the part of genes in the normalized first gene expression data.

3. The method according to claim 1 or 2, in, The step of processing the masked gene expression data and auxiliary information in the sample using the prediction model to be trained includes: splicing the masked gene expression data in the sample with the auxiliary information as the input feature tensor corresponding to the sample; or splicing the feature tensor corresponding to the masked gene expression data in the sample with the feature tensor corresponding to the auxiliary information as the input feature tensor corresponding to the sample; and The input feature tensor is processed using the prediction model to be trained.

4. The method according to claim 3, in, The prediction model to be trained includes an encoder network and a decoder network, and Wherein, the processing of the input feature tensor using the prediction model to be trained includes: encoding, using the encoder network, a first tensor corresponding to unmasked and non-zero counts in the input feature tensor to obtain an encoded tensor; Replacing the first tensor in the input feature tensor with the encoded tensor to obtain an encoded input feature tensor; and The encoded input feature tensor is decoded using the decoder network.

5. The method according to claim 1 or 2, in, The prediction model to be trained includes an input layer, an output layer, and a plurality of intermediate layers between the input layer and the output layer, and Wherein, the process of processing the masked gene expression data and auxiliary information in the sample by using the prediction model to be trained includes: Using the masked gene expression data in the sample as an input feature tensor or using the feature tensor corresponding to the masked gene expression data in the sample as an input feature tensor, inputting the input layer and a first predetermined layer among the plurality of intermediate layers to obtain an intermediate feature tensor, wherein the number of the first predetermined layers is greater than or equal to 0; splicing the auxiliary information with the intermediate feature tensor to obtain a spliced ​​intermediate feature tensor; and The concatenated intermediate feature tensor is input into a second predetermined layer among the multiple intermediate layers and the output layer, where the second predetermined layer is different from the first predetermined layer.

6. The method according to any one of claims 1 to 5, in, The prediction model to be trained is one of an encoder-decoder network, a decoder network, and a multi-layer perceptron.

7. The method according to claim 6, It is characterized in that The encoder includes M layers of encoding units, the decoder includes N layers of decoding units, and the value of M is greater than the value of N.

8. The method according to claim 7, It is characterized in that Each layer of encoding units of the encoder includes a multi-head attention unit and a forward propagation unit, and each layer of decoding units of the decoder includes a forward propagation unit and also includes a linear attention unit or a sparse attention unit.

9. The method according to any one of claims 1 to 8, in, The downsampled first gene expression data is obtained by downsampling the first gene expression data using a statistical sampling algorithm.

10. A method for correcting gene expression data, include: Acquiring current gene expression data and current auxiliary information, wherein the current gene expression data includes counts of different genes measured at an actual sequencing depth, and the current auxiliary information includes an expected first total count, the expected first total count is used to characterize an expected sequencing depth, and the expected first total count is greater than or equal to the sum of the counts of each gene in the current gene expression data; and The current gene expression data and the current auxiliary information are processed by at least part of the network layers in the prediction model trained according to the training method described in any one of claims 1 to 9 to obtain a correction value of the current gene expression data at the expected sequencing depth or an intermediate processing result of the correction value.

11. The method according to claim 10, in, The current auxiliary information also includes a current second total count, where the current second total count is the sum of the counts of each gene in the current gene expression data; Wherein, the obtaining of current gene expression data and current auxiliary information comprises: obtaining normalized current gene expression data and the current auxiliary information, wherein the normalized current gene expression data is obtained by normalizing the current gene expression data, and Wherein, processing the current gene expression data and the current auxiliary information using at least part of the network layers in the prediction model trained according to the training method according to any one of claims 1 to 9 to obtain a correction value of the current gene expression data at the expected sequencing depth or an intermediate processing result of the correction value includes: The normalized current gene expression data and the current auxiliary information are processed by at least part of the network layers in the prediction model trained according to the training method described in any one of claims 1 to 9 to obtain a correction value of the current gene expression data at the expected sequencing depth or an intermediate processing result of the correction value.

12. A downstream task execution method, include: Acquire input data, wherein the input data comprises i) a correction value obtained by the method according to claim 10 or 11; or ii) an intermediate processing result of the correction value in the method according to claim 10 or 11; or iii) a preprocessing result obtained by preprocessing the correction value obtained by the method according to claim 10 or 11; or iv) a preprocessing result obtained by preprocessing the intermediate processing result of the correction value obtained by the method according to claim 10 or 11; and The input data of the downstream task algorithm is used to obtain downstream task results, wherein the downstream tasks include cell classification tasks, disturbance prediction tasks or drug response prediction tasks.

13. A training device for a prediction model, include: A first unit is used to obtain multiple samples, wherein each sample of the multiple samples includes first gene expression data, masked gene expression data, and auxiliary information, the first gene expression data includes counts of different genes in a single cell measured at a first sequencing depth, the masked gene expression data is obtained by downsampling the first gene expression data and masking counts of some genes in the downsampled first gene expression data, the downsampled first gene expression data simulates counts of different genes in a single cell measured at a second sequencing depth lower than the first sequencing depth, and the auxiliary information includes a first total count, which is the sum of counts of each gene in the first gene expression data; A second unit is configured to, for each sample of the plurality of samples: Using the prediction model to be trained to process the masked gene expression data and auxiliary information in the sample to obtain a predicted value of each gene count at a first sequencing depth corresponding to the sample; Determining a loss value corresponding to the sample according to the predicted value and the counts corresponding to the part of genes in the first gene expression data; and The third unit is used to update the prediction model to be trained according to the loss value corresponding to each sample in the multiple samples.

14. A correction device for gene expression data, include: A fourth unit is used to obtain current gene expression data and current auxiliary information, wherein the current gene expression data includes counts of different genes measured at an actual sequencing depth, and the current auxiliary information includes an expected first total count, the expected first total count is used to characterize an expected sequencing depth, and the expected first total count is greater than or equal to the sum of the counts of each gene in the current gene expression data; and The fifth unit is used to process the current gene expression data and the current auxiliary information using at least part of the network layer in the prediction model trained by the training device according to claim 14, so as to obtain a correction value of the current gene expression data at the expected sequencing depth or an intermediate processing result of the correction value.

15. A downstream task execution device, include: a sixth unit, configured to obtain input data, wherein the input data includes i) a correction value obtained by the apparatus according to claim 14; or ii) an intermediate processing result of the correction value in the apparatus according to claim 14; or iii) a preprocessing result obtained by preprocessing the correction value obtained by the apparatus according to claim 14; or iv) a preprocessing result obtained by preprocessing the intermediate processing result of the correction value obtained by the apparatus according to claim 14; and The seventh unit is used to utilize the input data of the downstream task algorithm to obtain the downstream task result, wherein the downstream task includes a cell classification task, a disturbance prediction task or a drug response prediction task.

16. A computing device, include: a memory, a processor and a computer program stored on the memory, The processor is configured to execute the computer program to implement the steps of the method according to any one of claims 1 to 12.

17. A non-transitory computer readable storage medium having a computer program stored thereon, in, When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

18. A computer program product comprising computer instructions, in, When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.