Single cell data noise reduction method

Abstract

The invention discloses a single cell data noise reduction method, and relates to the field of cell data processing, the method is based on a DNA sequence and an RNA sequence to carry out modeling, predict chromatin accessibility and gene expression state, integrate paired single cell multi-omics data, and take the weight of a classifier as the cell low-dimensional characteristics of two omics, so that the noise reduction of the single cell data is realized. And cell heterogeneity is learned by using comparative learning, the real contour and biological information of the single cell data are described, and effective noise reduction is performed on the single cell data set.

Description

Technical Field

[0001] The present invention relates to the field of cell data processing, and in particular to a single cell data denoising method. Background Art

[0002] As an important branch of genetic technology, single-cell omics technology can reveal the genetic and expression differences between different cells through high-throughput sequencing and analysis of single cells, and gain in-depth understanding of information such as cell diversity, developmental processes, and disease mechanisms, thereby promoting the development of precision medicine. So far, many large-scale international cooperation projects have accumulated a large amount of data on human cells and genomes, which provides valuable resources for the development of single-cell omics technology.

[0003] As an important topic in modern bioinformatics, single-cell omics data analysis plays an important role in modern life sciences. Traditionally, biological research has mainly focused on a large number of homogeneous cell samples, ignoring the heterogeneity between cells. However, the emergence of single-cell omics technology has changed this situation. Parallel detection of cellular heterogeneity in genomes and proteomes can enable biomedical workers to have a deeper understanding of diseases and cells. However, the existence of "false zero" noise in current single-cell data will limit the further work of biomedical workers. Therefore, denoising single-cell data will effectively promote the development of corresponding disciplines. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, a single-cell data denoising method provided by the present invention realizes the correction of the counting matrix and completes the single-cell data denoising.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is: A single cell data denoising method is provided, which comprises the following steps: S1. Extract DNA sequences from scRNA-seq data and transcribe them into RNA sequences; Extract DNA sequences from scATAC-seq data; S2. Encode the transcribed RNA sequence and the DNA sequence extracted from the scATAC-seq data into one-hot encoding matrices, respectively, to obtain a DNA sequence encoding matrix and an RNA sequence encoding matrix; S3, splicing the DNA sequence coding matrix and the RNA sequence coding matrix to obtain a splicing matrix; S4, respectively obtaining the features of the DNA sequence encoding matrix, the features of the RNA sequence encoding matrix, and the features of the splicing matrix through stacked convolutional neural networks; S5, splicing the DNA sequence encoding matrix with its features to obtain a DNA fusion feature; splicing the RNA sequence encoding matrix with its features to obtain an RNA fusion feature; S6. The DNA fusion feature is fused with the feature of the splicing matrix through the trained gating network to obtain a DNA mixed feature; the RNA fusion feature is fused with the feature of the splicing matrix through the trained gating network to obtain an RNA mixed feature; S7, residually connect the DNA mixed features with the DNA sequence encoding matrix to obtain the sequence embedding of the scATAC-seq data; residually connect the RNA mixed features with the RNA sequence encoding matrix to obtain the sequence embedding of the scRNA-seq data; S8. Using the sequence embedding of scATAC-seq data as the input of the classifier in the trained gating network to obtain the predicted probability value of chromatin accessibility; using the sequence embedding of scRNA-seq data as the input of the classifier in the trained gating network to obtain the predicted probability value of gene expression status; S9. According to the predicted probability values ​​of chromatin accessibility and gene expression status, the counting matrix of scATAC-seq data and scRNA-seq data is modified to complete single-cell data noise reduction.

[0006] The beneficial effects of the present invention are: 1. This method builds models based on DNA sequences and RNA sequences, predicts chromatin accessibility and gene expression status simultaneously, and integrates paired single-cell multi-omics data. The classifier weights are used as the low-dimensional features of the two omics cells, and the cellular heterogeneity is learned by contrastive learning to improve the denoising effect of single-cell data.

[0007] 2. This method adopts a lightweight deep learning model, which avoids the embarrassing situation of the difficulty in training many current single-cell data fusion models based on encoder-decoder structures, and can also improve data processing efficiency.

[0008] 3. This method takes into account the impact of cell heterogeneity on the fusion of single-cell multi-omics data and uses comparative learning to more accurately characterize the single-cell characteristics in multi-omics data, laying the foundation for improving the noise reduction effect of single-cell data. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 is a schematic diagram of the process of this method; Figure 2 The performance comparison between the present method and the existing methods in the prediction of chromatin accessibility in the embodiment is shown; Figure 3 The performance comparison between the present method and the existing method in predicting gene expression status in the embodiment; Figure 4 The inference results in the 10X multiome PBMC data set of the present method in the embodiment; Figure 5 The inference results in the SNARE-seq data set of the present method in the embodiment; Figure 6 The inference results of the method in the SHARE-seq data set in the embodiment; Figure 7 The performance of the method in the embodiment in multi-omics data fusion; Figure 8 The performance of GLUE in multi-omics data fusion in the examples; Fig. 9 The performance of scBridge in multi-omics data fusion in the embodiment; Fig.10 The performance of scJoint in multi-omics data fusion in the embodiment; Fig.11 The cosine similarity between the two omics of the same cells in the 10X multiome PBMC dataset of the present method and the comparative method in the embodiment; Fig.12 The cosine similarity between the two omics of the same cells in the SNARE-seq dataset of the present method and the comparative method in the embodiment; Fig.13 The cosine similarity between the two omics of the same cells in the SHARE-seq dataset of the present method and the comparative method in the embodiment; Fig.14 The performance of the method in the embodiment in multi-omics data fusion on the SNARE-seq dataset; Fig.15 The performance of GLUE in multi-omics data fusion on the SNARE-seq dataset in the example; Fig.16 The performance of scBridge in multi-omics data fusion on the SNARE-seq dataset in the embodiment; Fig.17 The performance of scJoint in multi-omics data fusion on the SNARE-seq dataset in the embodiment; Fig.18 The denoising performance of scATAC-seq data and scRNA-seq data in the present method and the comparative method in the example on the 10X multiome PBMC dataset.

[0010] Fig.19The effect of the noise reduction of the present method in the embodiment on the correlation between the same cells in scATAC-seq data and scRNA-seq data; Fig. 20 The performance of the present method and the comparative method in terms of reasoning efficiency in the embodiments; Fig.21 The data profile of the scATAC-seq data before noise reduction in the embodiment; Fig. 22 The data profile of the scATAC-seq data in the embodiment after noise reduction by the method; Fig.23 The data profile of the scRNA-seq data before noise reduction in the embodiment; Fig.24 This is the data profile of the scRNA-seq data in the embodiment after noise reduction by this method. DETAILED DESCRIPTION

[0011] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0012] like Figure 1 As shown, the single cell data denoising method includes the following steps: S1. Extract DNA sequences from scRNA-seq data and transcribe them into RNA sequences; Extract DNA sequences from scATAC-seq data; S2. Encode the transcribed RNA sequence and the DNA sequence extracted from the scATAC-seq data into one-hot encoding matrices, respectively, to obtain a DNA sequence encoding matrix and an RNA sequence encoding matrix; S3, splicing the DNA sequence coding matrix and the RNA sequence coding matrix to obtain a splicing matrix; S4, respectively obtaining the features of the DNA sequence encoding matrix, the features of the RNA sequence encoding matrix, and the features of the splicing matrix through stacked convolutional neural networks; S5, splicing the DNA sequence encoding matrix with its features to obtain a DNA fusion feature; splicing the RNA sequence encoding matrix with its features to obtain an RNA fusion feature; S6. The DNA fusion feature is fused with the feature of the splicing matrix through the trained gating network to obtain a DNA mixed feature; the RNA fusion feature is fused with the feature of the splicing matrix through the trained gating network to obtain an RNA mixed feature; S7, residually connect the DNA mixed features with the DNA sequence encoding matrix to obtain the sequence embedding of the scATAC-seq data; residually connect the RNA mixed features with the RNA sequence encoding matrix to obtain the sequence embedding of the scRNA-seq data; S8. Using the sequence embedding of scATAC-seq data as the input of the classifier in the trained gating network to obtain the predicted probability value of chromatin accessibility; using the sequence embedding of scRNA-seq data as the input of the classifier in the trained gating network to obtain the predicted probability value of gene expression status; S9. According to the predicted probability values ​​of chromatin accessibility and gene expression status, the counting matrix of scATAC-seq data and scRNA-seq data is modified to complete single-cell data noise reduction.

[0013] The specific method of encoding into a one-hot encoding matrix is ​​as follows: the transcribed RNA sequence and the DNA sequence extracted from the scATAC-seq data are encoded into a sparse one-hot encoding matrix with a dimension of 5; the 5 dimensions correspond to adenine A, guanine G, cytosine C, thymine T and uracil U, respectively; the transcribed RNA sequence and the DNA sequence extracted from the scATAC-seq data are 1344-bp in length.

[0014] The stacked convolutional neural network consists of 7 layers of convolutional structures. The calculation expression of the first layer of convolutional structure is:

[0015] in is the output of the first layer of convolutional structure; , represents the DNA sequence encoding matrix, represents the RNA sequence encoding matrix, represents the concatenation matrix; and They are the weights and biases of the first layer of convolutional structure respectively; is the GLUE activation function; Represents a 1×1 convolution operation; represents batch normalization; The calculation expression of the second to seventh convolution structures is:

[0016] in For the The output of the layer convolution structure, when When the value is 7, is the feature corresponding to the input, ; For the The output of the layer convolutional structure; and Respectively The weights and biases of the layer convolutional structure; Represents a max pooling operation.

[0017] The expressions for obtaining DNA mixed features and RNA mixed features are:

[0018] in , DNA mixing characteristics, It is RNA mixed feature; , DNA fusion characteristics, It is the RNA fusion feature; and All are weights; and All are biased; is the feature of the concatenated matrix; is the softmax activation function; is the GLUE activation function.

[0019] The expressions for obtaining the sequence embeddings for scATAC-seq data and scRNA-seq data are:

[0020] in , Sequence embedding for scATAC-seq data, Sequence embedding for scRNA-seq data; and All are weights; and All are biased; , is the DNA sequence encoding matrix, Represents the RNA sequence encoding matrix.

[0021] The expression of the classifier in the trained gating network is:

[0022] in , is the predicted probability value of chromatin accessibility; is the predicted probability value of gene expression status; , Embedding for processing sequences The weight of the classifier, Embedding for processing sequences The weight of the classifier; , Embedding for processing sequences The bias of the classifier, Embedding for processing sequences The bias of the classifier; is the sigmoid activation function.

[0023] The specific method for training the gated network includes the following sub-steps: A1. Randomly initialize the weights and biases in the gating network. As a cell expression matrix for scATAC-seq data, Each vector in represents the feature vector of a cell; As the cell expression matrix for scRNA-seq data, Each vector in represents the feature vector of a cell; A2. Set the current Middle All feature vectors corresponding to the cell-like cells are averaged in the vector dimension, and the average result is used as the first Cell-like DNA cell prototype ; Set the current Middle All feature vectors corresponding to the cell-like cells are averaged in the vector dimension, and the average result is used as the first Cell-like RNA cell prototype ; Then we obtain two cell prototypes of cells used in the gating network training process of scATAC-seq data and scRNA-seq data; A3. Calculate the non-significant amount of gating network used in the training process of scATAC-seq data and scRNA-seq data. Cell-like prototypes and The distance vector ; The corresponding expression is:

[0024] in is the total number of cell types used in the gating network training process for scATAC-seq data and scRNA-seq data; is the temperature coefficient; Indicates Non Cell-like; It represents the exponential function with the natural constant e as the base; A4. Calculate the The distance vector between the cell prototypes currently corresponding to the class cell ; The corresponding expression is: ; A5. Construct a loss function to obtain the total loss value of the gating network, and update the weights and biases of the gating network by the total loss value until the loss function converges; the expression of the loss function is:

[0025]

[0026]

[0027]

[0028] in is the total loss value of the gating network; is the comparison learning loss value; The feature vector representing a single cell elements; The total number of elements in the feature vector representing a single cell; It represents the logarithmic function with the natural constant e as the base; is the binary cross entropy loss value corresponding to cell chromatin accessibility; Indicates True probability value of cell-like chromatin accessibility; Indicates The predicted probability value of chromatin accessibility corresponding to the cell class; Indicates The true probability value of the cell-like gene expression state; Indicates The predicted probability value of the cell-like gene expression state; represents the logarithmic function with base 10; is the binary cross entropy loss value corresponding to the cell gene expression state.

[0029] The specific method for correcting the count matrix of scATAC-seq data and scRNA-seq data is: using the predicted probability value of chromatin accessibility and the predicted probability value of gene expression status as the denoised data, completing the correction of the "false zero" counts existing in the count matrix of scATAC-seq data and the count matrix of scRNA-seq data, that is, completing the denoising of single-cell data.

[0030] In the specific implementation process, scRNA-seq data and scATAC-seq data contain a large amount of single-cell data. Therefore, during the training process, only part of the single-cell data is needed to train the gating network, and then the remaining single-cell data are denoised using this method.

[0031] In the specific implementation process, The initial distance between the two matrices in may be large, and the training makes The distance between the two matrices in the image is getting smaller and smaller. From the cellular level, the gap between the two groups of data for the same cell will also be very small. This process is called a fusion process.

[0032] In this embodiment, 10X multiome PBMC dataset, SNARE-seq dataset and SHARE-seq dataset were collected from the dataset processed by Gao Ge's team at Peking University. The 10X multiome PBMC dataset contains 19 single cell populations, 107,162 ATAC characteristic peak data and 29,083 gene expression data; the SNARE-seq dataset provides 22 single cell populations, 241,757 ATAC characteristic peak data and 28,918 gene expression data; and the SHARE-seq dataset contains 19 single cell populations, 340,314 ATAC characteristic peak data and 21,478 gene expression data.

[0033] In this example, the area under the receiver operating characteristic curve (auROC) was used to evaluate the performance of this method (CONTRACT) in the tasks of chromatin accessibility prediction and gene status prediction.

[0034] In this embodiment, the learning rate is set to 0.001, the training cycle (epoch) is set to 40, the data batch size is set to 128, and the adaptive moment estimation optimizer (Adam) is used to train CONTRCT. The training set, validation set, and test set are divided into 8:1:1 ratios.

[0035] For the DNA sequences retained in the above three standard data sets, this example compares the auROC values ​​of CONTRACT with those of scBasset, Basenji, and DeepSEA. For fair comparison, 1344-bp DNA and RNA sequences were used as input for all comparison methods.

[0036] like Figure 2 As shown in the figure, in terms of chromatin accessibility prediction, CONTRACT outperforms the existing prediction methods in all three datasets, especially in the 10X multiome PBMC dataset, where CONTRACT's advantage is more obvious, and its auROC is much higher than that of the other comparison methods.

[0037] like Figure 3 As shown in the figure, in terms of gene expression status prediction, CONTRACT still shows better prediction indicators than other methods, except that it is second only to Basenji in the SHARE dataset.

[0038] In this example, almost all benchmark methods perform much better on the SHARE-seq dataset than the other two datasets, probably because there is a lot of noise in the single-cell dataset, making it difficult for the model to fit the true distribution of the data, and the number of genes in the SHARE-seq dataset is much higher than the other two datasets, giving the model a better chance to learn the true data labels. In the case of limited data, the performance of CONTRACT shows that the use of multi-task learning and the mutual learning of chromatin accessibility prediction and gene state expression can help reduce the overfitting phenomenon of the model in scRNA-seq data processing and improve the performance of the model in the task of predicting gene expression state.

[0039] The high sparsity and "false zero" noise of the count matrices of scATAC-seq and scRNA-seq make it difficult for current methods to identify open chromatin regions to make further progress. In order to evaluate the robustness of CONTRACT in datasets with a lot of noise, this example randomly samples the non-zero counts in the count matrices of the three datasets to 80%, 60%, 40%, 20% and 10% of the original and retrains CONTRACT, and then completes the inference in the test set divided from the original dataset. Figure 4 , Figure 5 and Figure 6 The reasoning results of CONTRACT are presented, and the results show that even if the noise in the dataset is increased by sampling, the performance of CONTRACT is still slightly affected, which indicates that CONTRACT is very robust to noisy datasets.

[0040] like Figure 7 , Figure 8 , Fig. 9 and Fig.10 As shown, this embodiment qualitatively analyzes the performance of several comparative methods in multi-omics data fusion. CONTRACT and GLUE both well integrated scATAC-seq omics data and scRNA-seq omics data, eliminating the differences between the two omics data. However, scBridge and scJoint missed the fusion of some scRNA-seq and scATAC-seq omics data. The reason may be that, on the one hand, the operation of the scBridge and scJoint calculation methods depends on the alignment of the number of data in the non-cellular dimension. The experimental data used in this embodiment did not follow this feature. The number of characteristic peaks of scATAC-seq is much higher than the number of genes in scRNA-seq data. Therefore, some features of scATAC-seq data may be missed during the sampling process; on the other hand, as Fig.11 , Fig.12 , Fig.13 , Fig.14 , Fig.15 , Fig.16 and Fig.17 As shown in the figure, scBridge and scJoint use reliability modeling and "voting" to select cells for fusion, respectively. However, cells with high scores are not necessarily correct, which may cause the model to fuse wrong cells. This phenomenon is more obvious in the fusion of SNARE-seq datasets. After fusion, CONTRACT obtained the highest cosine similarity between omics, and GLUE also achieved high performance between omics. However, scBridge and scJoint have partially fused and unfused cell data, and the fusion performance between cells varies greatly.

[0041] In this embodiment, the predicted probability value of CONTRACT is used as the denoised data. This embodiment evaluates the denoising performance of CONTRACT on the 10Xmultiome PBMC dataset. Specifically, the denoised count matrix should be able to make the distances between the same cells, the same characteristic peaks, and the same genes closer. Therefore, this embodiment calculates the cell type ASW based on the denoised cell-characteristic peak count matrix and the cell-gene count matrix, which is an indicator for measuring cell similarity. In addition to CONTRACT, this embodiment uses the count matrix and the original count matrix generated by SCALE, MAGIC, and scBasset to calculate the cell type ASW respectively. Fig.18As shown in the figure, in the tasks of scATAC-seq data denoising and scRNA-seq data denoising, CONTRACT achieved the best indicators, followed by SCALE; MAGIC also achieved good performance in scATAC-seq data denoising.

[0042] This example further observes whether the distances between the same cells are more correlated after denoising by CONTRACT, and also measures such correlation by cosine similarity. Specifically, the cosine similarity between the vectors of the same cells when only denoising the scATAC-seq data, only denoising the scRNA-seq data, neither denoising, and both denoising is calculated, as shown in Figure 2. Fig.19 As shown in the figure, it was found that after both omics data were denoised, the correlation between the same cell vectors was higher, and if only one omics data was denoised, the cosine similarity between the same cell vectors was second, but higher than the cosine similarity of the data without denoising, indicating the effectiveness of CONTRACT for denoising both scATAC-seq data and scRNA-seq data.

[0043] As the amount of single-cell data continues to grow, the efficiency of the model becomes an important reference in the single-cell data analysis process. Since the denoising process runs through the reasoning process of CONTRACT and the above-mentioned comparison methods, Fig. 20 As shown in Figure 1, the running time of CONTRACT and the above methods for denoising scATAC-seq data is compared and recorded in detail in Table 1. Among them, scBasset has the fastest inference time of 190.01 seconds, and the running times of CONTRACT, SCALE, and MAGIC are 246.05, 883.00, and 344.57 seconds, respectively, indicating that CONTRACT can be used in the practical application of single-cell data denoising tasks like other comparison methods. However, all comparison methods can only denoise one type of omics data, while using multi-task learning, CONTRACT can denoise scATAC-seq data and scRNA-seq data at the same time. Deep learning such as CONTRACT, SCALE, and scBasset is expected to speed up its running efficiency as GPU hardware conditions improve, but the disadvantage is that the number of cells that can be analyzed is limited by the GPU video memory.

[0044] Table 1

[0045] like Fig.21 , Fig. 22 , Fig.23 and Fig.24As shown in the figure, the denoised and raw count matrices of scATAC-seq data and scRNA-seq data are visually compared. The denoised data can more accurately show the biological significance of single-cell data, making the outlines of the same cells closer. This shows that CONTRACT can show a more realistic distribution of single-cell data, and this performance is particularly evident in the denoising of scRNA-seq data.

[0046] In summary, in limited single-cell datasets, most existing single-cell chromatin accessibility prediction or gene status prediction models have serious overfitting phenomena. The present invention uses multi-task learning and comparative learning methods to alleviate this overfitting phenomenon when single-cell datasets are limited, improve the prediction performance of the model, effectively utilize cell heterogeneity, characterize the true outline and biological information of single-cell data, and effectively reduce the noise of single-cell datasets.

Claims

1. A single cell data denoising method, characterized in that: The following steps are involved: S1. Extract DNA sequences from scRNA-seq data and transcribe them into RNA sequences; Extract DNA sequences from scATAC-seq data; S2. Encode the transcribed RNA sequence and the DNA sequence extracted from the scATAC-seq data into one-hot encoding matrices, respectively, to obtain a DNA sequence encoding matrix and an RNA sequence encoding matrix; S3, splicing the DNA sequence coding matrix and the RNA sequence coding matrix to obtain a splicing matrix; S4, respectively obtaining the features of the DNA sequence encoding matrix, the features of the RNA sequence encoding matrix, and the features of the splicing matrix through stacked convolutional neural networks; S5, splicing the DNA sequence encoding matrix with its features to obtain a DNA fusion feature; splicing the RNA sequence encoding matrix with its features to obtain an RNA fusion feature; S6. The DNA fusion feature is fused with the feature of the splicing matrix through the trained gating network to obtain a DNA mixed feature; the RNA fusion feature is fused with the feature of the splicing matrix through the trained gating network to obtain an RNA mixed feature; S7, residually connect the DNA mixed features with the DNA sequence encoding matrix to obtain the sequence embedding of the scATAC-seq data; residually connect the RNA mixed features with the RNA sequence encoding matrix to obtain the sequence embedding of the scRNA-seq data; S8. Use the sequence embedding of scATAC-seq data as the input of the classifier in the trained gating network to obtain the predicted probability value of chromatin accessibility; The sequence embedding of scRNA-seq data is used as the input of the classifier in the trained gating network to obtain the predicted probability value of gene expression status; S9. According to the predicted probability values ​​of chromatin accessibility and gene expression status, the counting matrix of scATAC-seq data and scRNA-seq data is modified to complete single-cell data noise reduction.

2. The single cell data denoising method according to claim 1, characterized in that: The specific method of encoding into a one-hot encoding matrix is: The transcribed RNA sequences and the DNA sequences extracted from the scATAC-seq data were encoded as sparse one-hot encoding matrices with a dimension of 5; the 5 dimensions corresponded to adenine A, guanine G, cytosine C, thymine T, and uracil U, respectively; the transcribed RNA sequences and the DNA sequences extracted from the scATAC-seq data were 1344-bp in length.

3. The single cell data denoising method according to claim 1, characterized in that: The stacked convolutional neural network consists of 7 layers of convolutional structures. The calculation expression of the first layer of convolutional structure is: in is the output of the first layer of convolutional structure; , represents the DNA sequence encoding matrix, represents the RNA sequence encoding matrix, represents the concatenation matrix; and They are the weights and biases of the first layer of convolutional structure respectively; is the GLUE activation function; Represents a 1×1 convolution operation; represents batch normalization; The calculation expression of the second to seventh convolution structures is: in For the The output of the layer convolution structure, when When the value is 7, That is, the feature corresponding to the input, ; For the The output of the layer convolutional structure; and Respectively The weights and biases of the layer convolutional structure; Represents a max pooling operation.

4. The single cell data denoising method according to claim 1, characterized in that: The expressions for obtaining DNA mixed features and RNA mixed features are: in , DNA mixing characteristics, It is RNA mixed feature; , DNA fusion characteristics, It is the RNA fusion feature; and All are weights; and All are biased; is the feature of the concatenated matrix; is the softmax activation function; is the GLUE activation function.

5. The single cell data denoising method according to claim 4, characterized in that: The expressions for obtaining the sequence embeddings for scATAC-seq data and scRNA-seq data are: in , For sequence embedding of scATAC-seq data, Sequence embedding for scRNA-seq data; and All are weights; and All are biased; , is the DNA sequence encoding matrix, Represents the RNA sequence encoding matrix.

6. The single cell data denoising method according to claim 5, characterized in that: The expression of the classifier in the trained gating network is: in , is the predicted probability value of chromatin accessibility; is the predicted probability value of gene expression status; , Embedding for processing sequences The weight of the classifier, Embedding for processing sequences The weight of the classifier; , Embedding for processing sequences The bias of the classifier, Embedding for processing sequences The bias of the classifier; is the sigmoid activation function.

7. The single cell data denoising method according to claim 6, characterized in that: The specific method for training the gated network includes the following sub-steps: A1. Randomly initialize the weights and biases in the gating network. As a cell expression matrix for scATAC-seq data, Each vector in represents the feature vector of a cell; As the cell expression matrix of scRNA-seq data, Each vector in represents the feature vector of a cell; A2. Set the current Middle All feature vectors corresponding to the cell-like cells are averaged in the vector dimension, and the average result is used as the first Cell-like DNA cell prototype ; Set the current Middle All feature vectors corresponding to the cell-like cells are averaged in the vector dimension, and the average result is used as the first Cell-like RNA cell prototype ; Then we obtain two cell prototypes of cells used in the gating network training process of scATAC-seq data and scRNA-seq data; A3. Calculate the non-significant amount of gating network used in the training process of scATAC-seq data and scRNA-seq data. Cell-like prototypes and The distance vector ; The corresponding expression is: in is the total number of cell types used in the gating network training process for scATAC-seq data and scRNA-seq data; is the temperature coefficient; Indicates Non Cell-like; It represents the exponential function with the natural constant e as the base; A4. Calculate the The distance vector between the cell prototypes currently corresponding to the class cell ; The corresponding expression is: ； A5. Construct a loss function to obtain the total loss value of the gating network, and update the weights and biases of the gating network by the total loss value until the loss function converges; the expression of the loss function is: in is the total loss value of the gating network; is the comparison learning loss value; The feature vector representing a single cell elements; The total number of elements in the feature vector representing a single cell; It represents the logarithmic function with the natural constant e as the base; is the binary cross entropy loss value corresponding to cell chromatin accessibility; Indicates True probability value of cell-like chromatin accessibility; Indicates The predicted probability value of chromatin accessibility corresponding to the cell class; Indicates The true probability value of the cell-like gene expression state; Indicates The predicted probability value of the cell-like gene expression state; represents the logarithmic function with base 10; is the binary cross entropy loss value corresponding to the cell gene expression state.

8. The single cell data denoising method according to claim 1, characterized in that: The specific method for correcting the count matrix of scATAC-seq data and scRNA-seq data is as follows: The predicted probability values ​​of chromatin accessibility and gene expression status were used as denoised data to correct the "false zero" counts in the count matrix of scATAC-seq data and the count matrix of scRNA-seq data, thus completing the denoising of single-cell data.

Images