Data analysis system and method of gene regulation and control network based on deep regression algorithm

Abstract

The invention is suitable for the technical field of data analysis, and provides a data analysis system and method of a gene regulation and control network based on a deep regression algorithm, and the method comprises the following steps: converting gene identity and gene expression quantity values into vector representation; calculating an attention score between gene vector representations as an evaluation criterion of a relationship between genes; and predicting the specific gene expression quantity. By introducing the Gaussian layer, the anti-noise performance is enhanced, embedding and attention mechanisms are utilized, the complex regulation and control relation between genes is captured, individualized GRN prediction is achieved on the single cell level, the defects of an existing method in individualized treatment of cancer patients can be overcome, and the method has a good application prospect. And the gene expression quantity analysis and prediction in individual treatment of cancer patients are supported.

Description

Technical Field

[0001] The present invention belongs to the technical field of data analysis, and in particular to a data analysis system and method for a gene regulatory network based on a deep regression algorithm. Background Art

[0002] Gene expression in cancer patients plays a vital role in predicting disease progression and guiding drug selection. By analyzing changes in gene expression, we can reveal the function and role of specific genes in the disease. Single-cell RNA sequencing (scRNA-seq) technology provides detection of gene expression in individual cells, which can reflect individual differences in patients and heterogeneity within tumors. This provides detailed cellular level information for personalized analysis. Gene regulatory networks (GRNs) aim to reveal the regulatory relationship between transcription factors and target genes. Through accurate GRN predictions, we can better understand the mechanism of changes in gene expression and provide stronger decision support for personalized cancer treatment.

[0003] However, the current analysis and prediction of gene expression in personalized treatment of cancer patients, especially the modeling work based on single-cell RNA sequencing (scRNA-seq) data, still faces many technical difficulties. Traditional data processing and analysis methods are easily affected by noise when processing single-cell data, lack accurate capture of complex gene regulatory networks (GRNs), and have difficulty supporting causal inference. For example, most existing analysis methods can only process static data and cannot achieve dynamic modeling of gene expression at the single-cell level, thus limiting the accuracy of the model in predicting changes in specific gene expression in cancer patients. In addition, in the presence of a large amount of noise, the performance of traditional analysis methods is significantly reduced, which can easily lead to incorrect identification of gene regulatory relationships.

[0004] Although existing single-cell gene expression data analysis methods (such as Gaussian mixture models and differential expression analysis) can handle a certain degree of noisy data, they lack the ability to adapt to the heterogeneity and individual specificity of cancer patients. When faced with personalized treatment data, these methods are often unable to filter out non-biological noise in the data, resulting in unstable model inference results and difficulty in identifying complex gene regulatory relationships within tumors. In addition, although many GRN inference methods can simulate the relationship between some transcription factors and target genes, due to the difficulty in modeling nonlinear and multidimensional relationships between genes, they are often unable to achieve accurate gene prediction and regulatory network reconstruction at the single-cell level.

[0005] Although existing single-cell gene expression simulation tools and GRN inference tools take gene regulatory relationships into account to a certain extent, these methods lack causal structural support and are prone to distortion in noisy data. Especially in cancer research applications, common tumor samples have large individual differences and high noise intensity. Existing tools are difficult to support accurate GRN modeling for cancer individuals, resulting in unreliable results. Summary of the invention

[0006] The purpose of the embodiments of the present invention is to provide a data analysis method for a gene regulatory network based on a deep regression algorithm, aiming to solve the problems raised in the above-mentioned background technology.

[0007] The embodiment of the present invention is implemented as follows: a data analysis method for a gene regulatory network based on a deep regression algorithm comprises the following steps: Convert gene identity and gene expression values ​​into vector representations; Calculate the attention score between gene vector representations as a criterion for judging the relationship between genes; Predict the expression level of specific genes.

[0008] Preferably, the step of converting the gene identity and gene expression value into a vector representation specifically includes: Encodes genetic identity; Gaussian distribution is used to encode the gene expression values; Add the two parts of the encoding to get the final vector representation.

[0009] Preferably, the step of encoding the gene expression value using Gaussian distribution is specifically as follows: for a single gene expression value data g, a set of uniformly distributed Gaussian basis functions G are used to encode the distance, and each Gaussian basis function has its own mean , standard deviation : ; A single value g will be represented as: .

[0010] Preferably, the step of encoding the gene expression value using Gaussian distribution further includes: using a separate vector representation to additionally consider encoding of missing values.

[0011] Preferably, the step of calculating the attention scores between gene vector representations as a criterion for judging the relationship between genes is specifically as follows: for each gene, the attention score between the gene and other genes is calculated. After all genes are calculated, these attention scores are sorted from high to low, and it is considered that there is a gene regulatory relationship between the genes corresponding to the top n attention scores.

[0012] Another object of an embodiment of the present invention is to provide a data analysis system for a gene regulatory network based on a deep regression algorithm, which is used to implement the data analysis method for a gene regulatory network based on a deep regression algorithm, including: The encoding layer module is used to convert gene identity and gene expression values ​​into vector representations; The self-attention layer module is used to calculate the attention scores between gene vector representations as a criterion for judging the relationship between genes; The prediction layer module is used to predict the specific gene expression level.

[0013] Preferably, the coding layer module includes: Gene identity encoding layer unit, used to encode gene identity; The expression value Gaussian encoding layer unit is used to encode the gene expression value using Gaussian distribution; The adding unit is used to add the two parts of the code to obtain the final vector representation.

[0014] Preferably, the parameters of the self-attention layer module are: 8 self-attention heads, 1 self-attention layer, and 64 vector encoding length of the self-attention layer.

[0015] Preferably, it also includes: a model training module, which is used to randomly divide the samples into a training set, a validation set and a test set. The training set is responsible for training the model and learning the potential gene regulatory network GRN. The validation set is responsible for preventing the model from overfitting and is used to select the optimal model. The test set is responsible for testing the performance of the model in predicting gene expression.

[0016] The data analysis method of the gene regulatory network based on the deep regression algorithm provided in the embodiment of the present invention enhances the noise resistance by introducing the Gaussian layer, captures the complex regulatory relationship between genes by using the embedding and attention mechanism, and realizes individualized GRN prediction at the single-cell level. It can overcome the defects of the existing methods in the personalized treatment of cancer patients, and further support the analysis and prediction of gene expression in the personalized treatment of cancer patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of a data analysis method for a gene regulatory network based on a deep regression algorithm provided in an embodiment of the present invention; Figure 2 A block diagram of a data analysis system for a gene regulatory network based on a deep regression algorithm provided by an embodiment of the present invention; Figure 3 A block diagram of a coding layer module and a coding process diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0019] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0020] Example 1: Data analysis method of gene regulatory network based on deep regression algorithm, such as Figure 1 The block diagram of the analysis system is shown in Figure 1 As shown, the following steps are included: First, the data will pass through the system's encoding layer module, which converts the gene identity and gene expression value into a vector representation; This vector representation is used as the input of the self-attention layer module. The model implementation of the self-attention layer is the same as that of BERT (Bidirectional Encoder Representations from Transformers). The context information is comprehensively considered. The self-attention heads in the self-attention layer module parameters are set to 8, the number of self-attention layers is set to 1, and the vector encoding length of the self-attention layer is set to 64. By inputting the output of the self-attention layer module into the prediction layer module, the specific gene expression level can be predicted.

[0021] Among them, the coding layer module, such as Figure 2 As shown, the encoding process is as follows: Encodes genetic identity; Multiple Gaussian distributions are used to encode the gene expression values. Specifically, for a single gene expression value g, a set of uniformly distributed Gaussian basis functions G are used to encode the distance. Each Gaussian basis function has its own mean. , standard deviation : , and finally the single value g will be represented as , and additionally consider the encoding of missing values ​​using a separate vector representation, which is a learnable parameter; After encoding the gene identity and gene expression, the two parts are added together to obtain the final vector representation as the input of the self-attention layer.

[0022] In addition, the data analysis system also includes a model training module. Specifically, during the training process, a single sample will be randomly set as a missing value with a probability of 50%, so that the model can predict the missing value during the training process, so that the model can reduce the error of predicting the missing value; The N*M gene expression matrix of the original analysis data can be randomly divided into 80%, 10%, and 10% as training set, validation set, and test set, respectively. The training set is responsible for training the model and learning the potential gene regulatory network GRN. The validation set is responsible for preventing the model from overfitting and is used to select the optimal model. The test set is responsible for testing the performance of the model in predicting gene expression. The parameters of this module are: set the training to 100 rounds, the number of samples in each iteration in each round is 32, the learning rate is 0.001, and the number of warm-up iterations of the learning rate is set to 5000.

[0023] Example 2: Analysis process of the data analysis system of the gene regulatory network based on the deep regression algorithm: For the gene x of interest, 50 samples are randomly selected from the training data; The trained optimal model is used to predict the gene expression values ​​in these 50 samples. The self-attention score of gene x in the self-attention layer of the model for other genes during the prediction process represents the regulatory relationship of other genes to gene x. The larger the value, the stronger the regulatory relationship. After sorting the self-attention scores, the top 100 genes are selected as the regulatory genes of the gene, that is, the gene regulatory network of the sample is constructed.

[0024] Example 3: Analysis was performed in three single-cell data GSD, HSC, and VSC constructed from the public data set BEELINE. These three single-cell data GSD, HSC, and VSC have 20,000 sample data respectively, and 18,000 sample data are randomly divided as training data, 1,000 data are used as verification data, and 1,000 data are used as test data. They have 79 genes, 30 genes, and 15 genes respectively, all of which provide verified gene regulatory relationships. AUROC and AUPR can be used to judge the accuracy of gene regulatory relationship prediction, and MSE can be used to judge the prediction error of gene expression. The data analysis system provided in Example 1 is compared with GENIE3, GRNBoost2, and scGeneRAI in terms of constructing gene relationships, and is also compared with scGeneRAI in terms of predicting gene expression. The results are shown in Tables 1 and 2. It can be seen from the following table that the model captures the complex regulatory relationship between genes and realizes individualized gene expression prediction at the single-cell level:

[0025]

[0026] In summary, the data analysis method of the gene regulatory network based on the deep regression algorithm provided in the embodiment of the present invention introduces a Gaussian layer to enhance the noise processing capability of the system. At the same time, through the combination of the gene embedding layer and the attention mechanism, a more accurate GRN inference effect is achieved in capturing the complex interaction relationship between genes. It can infer individual-specific GRNs based on single-cell data, support personalized treatment, have high versatility, and can adapt to different types of cancer samples, thus having wide application potential in precision medicine.

[0027] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A data analysis method for gene regulatory networks based on a deep regression algorithm, characterized in that: The following steps are involved: Convert gene identity and gene expression values ​​into vector representations; Calculate the attention score between gene vector representations as a criterion for judging the relationship between genes; Predict specific gene expression levels; The step of converting the gene identity and gene expression value into a vector representation specifically includes: Encodes genetic identity; Gaussian distribution is used to encode the gene expression values; Add the two parts of the code to get the final vector representation; The step of encoding the gene expression value using Gaussian distribution is specifically as follows: for a single gene expression value data g, a set of uniformly distributed Gaussian basis functions G are used to encode the distance, each Gaussian basis function having its own mean , standard deviation : ； A single value g will be represented as: ; The step of encoding the gene expression value using Gaussian distribution also includes: using a separate vector representation to additionally consider encoding missing values.

2. The method for analyzing gene regulatory network data based on deep regression algorithm according to claim 1, characterized in that: The step of calculating the attention scores between gene vector representations as a criterion for judging the relationship between genes is specifically as follows: for each gene, the attention scores between the gene and other genes are calculated. After all genes are calculated, the attention scores are sorted from high to low, and it is considered that there is a gene regulatory relationship between the genes corresponding to the top n attention scores.

3. A data analysis system for a gene regulatory network based on a deep regression algorithm, used to implement the data analysis method for a gene regulatory network based on a deep regression algorithm according to claim 1 or 2, characterized in that: include: The encoding layer module is used to convert gene identity and gene expression values ​​into vector representations; The self-attention layer module is used to calculate the attention scores between gene vector representations as a criterion for judging the relationship between genes; The prediction layer module is used to predict the specific gene expression level.

4. The data analysis system for gene regulatory networks based on deep regression algorithm according to claim 3, characterized in that: The coding layer module comprises: Gene identity encoding layer unit, used to encode gene identity; The expression value Gaussian encoding layer unit is used to encode the gene expression value using Gaussian distribution; The adding unit is used to add the two parts of the code to obtain the final vector representation.

5. The data analysis system of gene regulatory network based on deep regression algorithm according to claim 3, characterized in that: The parameters of the self-attention layer module are: 8 self-attention heads, 1 self-attention layer, and 64 vector encoding length of the self-attention layer.

6. The data analysis system of gene regulatory network based on deep regression algorithm according to claim 4, characterized in that: Also includes: The model training module is used to randomly divide the samples into training set, validation set and test set. The training set is responsible for training the model and learning the potential gene regulatory network GRN. The validation set is responsible for preventing the model from overfitting and is used to select the optimal model. The test set is responsible for testing the performance of the model in predicting gene expression.

Images