Method for marking renal clear cells by sialyltransferase gene signature for non-disease diagnosis purpose and application
By combining sialic acid transferase gene signature and random forest algorithm, candidate gene sets are generated and predictive models are constructed, which solves the systematic and accurate problems of gene analysis in traditional methods, and achieves more efficient renal transparent cell marking and prediction effects .
Patent Information
- Application Number
- CN202411435067.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-15
- Publication Date
- 2024-12-27
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional methods lack systematicity and comprehensiveness in cell gene analysis, it is difficult to identify complex gene interactions and deep information, and it is difficult to build accurate predictive models, resulting in poor gene signature recognition effect and limited model generalization ability.
The sialic acid transferase gene signature method was used to generate candidate gene sets through fold change and t test, and a random forest algorithm was used to construct feature selection and prediction models, select the most distinguished gene combination as the best signature, and verify the model through fold change accuracy.
It significantly improves the accuracy of gene signature and generalization ability of the model, ensures the quality of candidate genes and the reliability of predicted results, and provides a more innovative analytical paradigm.
Smart Images
Figure CN119207561A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of cell gene analysis, and in particular to a method and application of using a sialyltransferase gene signature for non-disease diagnosis purposes to label renal clear cells. Background Art
[0002] First, traditional methods are often limited to analyzing the expression differences of a single or a few preset genes. They lack systematicity and comprehensiveness and are prone to missing important genetic information, resulting in one-sided and incomplete research results.
[0003] Secondly, traditional methods have limited capabilities when processing high-dimensional genetic data, and it is difficult to effectively identify and analyze complex gene interactions and internal patterns, and it is impossible to fully tap the deep information contained in the data. Furthermore, traditional methods cannot automatically learn the inherent laws in the data, and it is difficult to accurately evaluate the contribution of each gene and its combination to classification, so they are not effective in identifying the most discriminative gene signatures. In addition, traditional methods often use simple statistical methods when building predictive models, which are difficult to cope with the complexity and noise of biological data, resulting in limited generalization ability and prediction accuracy of the model.
[0004] Finally, traditional methods lack the ability to process gene expression data comprehensively, deeply and efficiently, and cannot provide reliable and innovative analysis paradigms for the study of renal clear cells and other cell types. These problems seriously restrict the application effect and research value of traditional methods in gene expression analysis. Summary of the invention
[0005] The purpose of the present invention is to provide a method and application of using a sialyltransferase gene signature for labeling renal clear cells for non-disease diagnosis purposes, so as to solve the problems raised in the above-mentioned background technology.
[0006] To achieve the above objectives, one of the objectives of the present invention is to provide a method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes, the method comprising the following steps: S1. Collect cell samples, wherein the cell samples include cell sample ID, sample type and sialyltransferase gene expression data; divide the cell samples into a cell training set and a cell validation set, and select renal clear cell samples and normal renal cell samples in the cell training set as the total gene set; S2. Based on the sialyltransferase gene expression data of renal clear cell samples and the sialyltransferase gene expression data of normal renal cells in the total gene set, the gene differences between the two were compared using fold change and t-test methods, and a candidate gene set was generated; S3. Based on the candidate gene set, a feature selection model is constructed using the random forest algorithm. Based on the constructed feature selection model, the most discriminative gene combination is selected as the best gene signature; S4. Select the corresponding sialyltransferase gene expression data in the total gene set according to the best gene signature to generate a gene signature data set; and construct a prediction model using a random forest algorithm based on the gene signature data set; In addition, renal clear cell samples and normal renal cell samples were selected from the sample types in the cell validation set as the validation gene set, and the sample type of the validation gene set was predicted using the constructed prediction model; based on the predicted sample type, the prediction model was verified using the fold change.
[0007] As a further improvement of the technical solution, the use of fold change and t-test method to compare the gene differences between the two and generate a candidate gene set specifically includes: For the expression value corresponding to each gene type of sialyltransferase gene expression data in the total gene set, calculate its mean value and standard deviation in renal clear cell samples and normal renal cell samples; and calculate the ratio of the mean value of each gene type in renal clear cell samples and normal renal cell samples to generate the fold change results of each gene type in renal clear cell samples and normal renal cell samples; The t statistic was calculated based on the mean and standard deviation of each gene type in renal clear cell samples and normal renal cell samples, and the formula is as follows: ,in and Respectively represent the average values of gene types in renal clear cell samples and normal renal cell samples, and The standard deviation of the gene types in the renal clear cell samples and normal renal cell samples, respectively, and The number of samples with the respective genotypes in samples representing renal clear cells and samples representing normal renal cells; Calculating degrees of freedom , the formula is According to the t statistic and degrees of freedom, the t distribution table is used to obtain the Value, where The value is a probability value used in statistics to quantify the degree of difference between the observed data and the null hypothesis; set up The threshold of the value and the threshold of the fold change are set; the screening The gene types with the best values and fold change thresholds were selected as candidate gene sets.
[0008] As a further improvement of the technical solution, the construction of the feature selection model using the random forest algorithm specifically includes: The candidate gene set was organized into a table, where the rows represented different cell sample IDs and the columns included sialyltransferase gene type, expression value, and sample type; Using the combination principle, the candidate gene sets are used to generate all combination sets; Select a gene combination from the combination set in the order of the list as the current gene combination, extract the corresponding expression value and sample type in the table according to the current gene combination, and use it as the training set of the feature selection random forest algorithm; Samples are randomly extracted with replacement from a training set of a feature selection random forest algorithm, and each node of a decision tree in the feature selection random forest is constructed by randomly selecting a features as candidates from the current gene combination in the training set of the feature selection random forest algorithm, where a is the square root of the total number of gene types in the current gene combination; recursively construct until a stopping condition is reached, and the stopping condition includes but is not limited to reaching a preset maximum tree depth and the number of samples in the node is less than a preset minimum number.
[0009] As a further improvement of the technical solution, the most discriminative gene combination is selected as the best gene signature according to the constructed feature selection model, specifically including: A decision tree is constructed for each gene combination in the combination set to form a feature selection model. The average importance score of each gene combination is calculated based on the decision tree corresponding to the gene. The average importance is based on the degree to which the gene reduces impurity in the tree. For each gene in the combination, the average importance score in all decision trees is calculated. The average importance scores of all genes in the combination are added and averaged to obtain the average importance score of the combination. All gene combinations were ranked in descending order according to the average importance score of each combination, and the combination with the highest score was selected as the best gene signature that could effectively distinguish sialyltransferases from renal clear cells.
[0010] As a further improvement of the technical solution, the construction of the prediction model using the random forest algorithm specifically includes: According to the gene combination in the best gene signature, the expression data and sample type of the sialyltransferase gene corresponding to the gene combination are extracted from the total gene set and used as the gene signature data set; According to the gene signature dataset, the prediction model is constructed using the random forest algorithm. Each prediction decision tree is randomly sampled with replacement. At each branch node, a subset of the gene signature dataset is randomly selected for splitting, and the process is repeated until the prediction model converges.
[0011] As a further improvement of the technical solution, the prediction of sample type for the validation gene set using the constructed prediction model specifically includes: For the validation gene set, the gene expression values corresponding to the best gene signature are extracted and input into each prediction decision tree in the prediction model. Each prediction decision tree gives a prediction result. The prediction results of all prediction decision trees are summarized, and the majority voting method is used to determine the final predicted sample type.
[0012] As a further improvement of the technical solution, the predicted validation gene set is validated by using fold changes, specifically including: For the validation gene set and the predicted sample type, the average value of each gene in the validation gene set in the predicted sample type is calculated, and the average value is compared with the average value of the actual sample type to generate the validation fold change; The validation fold change is compared with the set fold change threshold. If the threshold requirement is met, its prediction model is output as the final model, which will predict the sample type of the new cell based on the gene type in the best gene signature.
[0013] The second object of the present invention is to provide a method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes in cell gene analysis.
[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. The multiple applications of fold changes in the labeling method and application of a sialyltransferase gene signature for non-disease diagnosis purposes for renal clear cells have produced significant beneficial effects; first, when generating a candidate gene set, the use of fold changes in combination with the t-test method can effectively identify sialyltransferase genes with significant differences in expression levels between renal clear cells and normal renal cells; this dual screening mechanism not only considers statistical significance, but also pays attention to biological significance, ensuring that the selected candidate genes have high discrimination and reliability; secondly, the fold changes are applied again in the prediction model verification stage, and the accuracy and stability of the model are further verified by comparing the ratio of the mean value of gene expression in the predicted sample type and the actual sample type; this multiple application method fully utilizes the advantages of fold changes in gene expression difference analysis, which not only improves the quality of candidate genes in the initial screening stage, but also ensures the reliability of the prediction results in the final verification stage, thereby greatly improving the accuracy and practicality of the entire labeling method.
[0015] 2. The multiple applications of the random forest algorithm in the method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes played a key role in this method and produced significant beneficial effects. First, in the feature selection stage, the random forest algorithm was used to construct a feature selection model, and by generating all possible gene combinations and evaluating the average importance score of each combination, the most discriminative gene combination was finally selected as the best gene signature. This process fully utilized the advantages of the random forest algorithm in processing high-dimensional data and evaluating feature importance, and effectively screened out the optimal combination from a large number of candidate genes. Secondly, in the prediction model construction stage, the random forest algorithm was applied again to construct a prediction model based on the best gene signature data set. This ensemble learning-based method can effectively handle the complexity and noise of biological data, improve the generalization ability of the model and the prediction accuracy. Finally, in the verification stage, the constructed prediction model was used to predict the sample type of the verification gene set, and further verified in combination with the fold change, ensuring the reliability and practicality of the model. This multiple application fully utilized the advantages of the random forest algorithm in feature selection, model construction and prediction, and greatly improved the efficiency and accuracy of the entire labeling method. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 It is a schematic diagram of the method steps of the present invention. DETAILED DESCRIPTION
[0017] The following will be combined with the accompanying drawings in the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0018] See also Figure 1 One of the purposes of this embodiment is to provide a method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes, comprising the following method steps: S1. Collect cell samples, where the cell samples include cell sample ID, sample type and sialyltransferase gene expression data; divide the cell samples into a cell training set and a cell validation set, and select renal clear cell samples and normal renal cell samples from the sample types in the cell training set as the total gene set.
[0019] S2. Based on the sialyltransferase gene expression data of renal clear cell samples and normal renal cells in the total gene set, the gene differences between the two were compared using fold change and t-test methods, and a candidate gene set was generated.
[0020] S3. Based on the candidate gene set, the random forest algorithm is used to construct a feature selection model. Based on the constructed feature selection model, the most discriminative gene combination is selected as the best gene signature.
[0021] S4. According to the best gene signature, the corresponding sialyltransferase gene expression data in the total gene set is selected to generate a gene signature data set; based on the gene signature data set, a prediction model is constructed using a random forest algorithm.
[0022] And select renal clear cell samples and normal renal cell samples from the sample types in the cell validation set as the validation gene set, and use the constructed prediction model to predict the sample type of the validation gene set; according to the predicted sample type, use the fold change to verify the prediction model.
[0023] Specifically include: In S1, cell samples are collected through a public gene expression database; the cell samples include cell sample ID, sample type and sialyltransferase gene expression data; the cell samples are divided into a cell training set and a cell validation set, and the sample types of renal clear cell samples and normal renal cell samples in the cell training set are selected as the total gene set; the renal clear cell samples represent tumor tissues from patients with renal clear cell carcinoma, and the normal renal cell samples represent renal tissues from healthy individuals or normal tissues adjacent to the cancer from cancer patients; the sialyltransferase gene expression data include multiple gene types and expression values, such as ST3GAL (neuraminic acid α-2,3-sialyltransferase), ST6GAL (neuraminic acid α-2,6-sialyltransferase), ST6GALNAC (neuraminic acid α-2,6-sialyltransferase N-acetylgalactosamine) and ST8SIA (neuraminic acid α-2,8-sialyltransferase), etc., and the expression values are expressed in numerical form.
[0024] S2. Based on the sialyltransferase gene expression data of renal clear cell samples and normal renal cells in the total gene set, the gene differences between the two were compared using fold change and t-test methods, and a candidate gene set was generated, including: For the expression values corresponding to the gene types of each sialyltransferase gene expression data in the total gene set, the mean and standard deviation in renal clear cell samples and normal renal cell samples were calculated; and the ratio of the mean values of each gene type in renal clear cell samples and normal renal cell samples was calculated to generate the fold change results of each gene type in renal clear cell samples and normal renal cell samples.
[0025] The t statistic was calculated based on the mean and standard deviation of each gene type in renal clear cell samples and normal renal cell samples, and the formula is as follows: ,in and Respectively represent the average values of gene types in renal clear cell samples and normal renal cell samples, and The standard deviation of the gene types in the renal clear cell samples and normal renal cell samples, respectively, and The number of samples with the respective genotypes in samples representing renal clear cells and samples representing normal renal cells; Calculating degrees of freedom , the formula is According to the t statistic and degrees of freedom, the t distribution table is used to obtain the Value, where The value is a probability value used in statistics to quantify the degree of difference between the observed data and the null hypothesis; set up The threshold of the value is usually 0.05 or 0.01; and the threshold of the fold change is set; the screening is performed to meet the The gene types with the best values and fold change thresholds were selected as candidate gene sets.
[0026] S3. Based on the candidate gene set, a feature selection model was constructed using the random forest algorithm. The most discriminative gene combination was selected based on the constructed feature selection model as the best gene signature that can effectively distinguish the sialyltransferase of renal clear cells, including: The candidate gene set was organized into a table where the rows represented different cell sample IDs and the columns included sialyltransferase gene type, expression value, and sample type (kidney clear cell or normal kidney cell).
[0027] Using the combinatorial principle, the candidate gene set is used to generate all combination sets. For example, all combination sets of 8 genes need to be generated. All combination sets include single genes, combinations of two genes, combinations of three genes, and up to combinations of all 8 genes. There will be a total of 255 different combinations (2^8-1), and each combination represents a potential gene signature.
[0028] Select a gene combination from the combination set in the order of the list as the current gene combination, extract the corresponding expression value and sample type in the table according to the current gene combination, and use it as the training set for the feature selection random forest algorithm.
[0029] Samples are randomly selected with replacement from the training set of the feature selection random forest algorithm, and each node of the decision tree in the feature selection random forest is constructed, by randomly selecting a features as candidates from the current gene combination in the training set of the feature selection random forest algorithm, where a is the square root of the total number of gene types in the current gene combination, for example, the current gene combination contains 9 gene types, then a=3; recursively construct until a stopping condition is reached, and the stopping condition includes but is not limited to reaching a preset maximum tree depth and the number of samples in the node is less than a preset minimum number.
[0030] A decision tree is constructed for each gene combination in the combination set to form a feature selection model. The average importance score is calculated based on the decision tree corresponding to each gene combination, where the average importance is based on the degree to which the gene reduces impurity (such as Gini impurity or information gain) in the tree. For each gene in the combination, the average importance score in all decision trees is calculated. The average importance scores of all genes in the combination are added and averaged to obtain the average importance score of the combination.
[0031] All gene combinations were ranked in descending order according to the average importance score of each combination, and the combination with the highest score was selected as the best gene signature that could effectively distinguish sialyltransferases from renal clear cells.
[0032] S4. Select the corresponding sialyltransferase gene expression data in the total gene set according to the best gene signature to generate a gene signature data set; construct a prediction model using a random forest algorithm based on the gene signature data set, specifically including: According to the gene combination in the best gene signature, the expression data of the sialyltransferase genes corresponding to the gene combination and the sample type are extracted from the total gene set and used as the gene signature data set.
[0033] According to the gene signature dataset, the prediction model is constructed using the random forest algorithm. Each prediction decision tree is randomly sampled with replacement. At each branch node, a subset of the gene signature dataset is randomly selected for splitting, and the process is repeated until the prediction model converges.
[0034] The sample types in the cell validation set are kidney clear cell samples and normal kidney cell samples as the validation gene set, where the sample types in the validation gene set are actual cell types; and the constructed prediction model is used to predict the sample type of the validation gene set, specifically including: For the validation gene set, the gene expression values corresponding to the best gene signature were extracted and input into each prediction decision tree in the prediction model. Each prediction decision tree gave a prediction result. The prediction results of all prediction decision trees were summarized, and the majority voting method was used to determine the final predicted sample type (renal clear cells or normal renal cells).
[0035] For the predicted validation gene set, the fold change is used to validate it, including: For the validation gene set and the predicted sample type, the average value of each gene in the validation gene set in the predicted sample type is calculated, and the average value is compared with the average value of the actual sample type to generate the validation fold change; The validation fold change is compared with the set fold change threshold. If the threshold requirement is met, its prediction model is output as the final model, which will predict the sample type of the new cell based on the gene type in the best gene signature.
[0036] The model was validated using cell validation sets and fold changes to ensure the reliability of the prediction results; this method integrated a variety of advanced analytical techniques to improve the accuracy and efficiency of renal clear cell markers, providing an innovative analytical paradigm for related research.
[0037] The following simplified example shows the entire process, including S1 to S4, as follows: S1. Cell sample data: Sample ID1, renal clear cells, ST3GAL1:100, ST3GAL2:150, ST6GAL1:200, ST6GAL2:180; Sample ID2, renal clear cells, ST3GAL1:110, ST3GAL2:160, ST6GAL1:190, ST6GAL2:170; Sample ID3, normal kidney cells, ST3GAL1:50, ST3GAL2:70, ST6GAL1:80, ST6GAL2:90; Sample ID4, normal kidney cells, ST3GAL1:60, ST3GAL2:80, ST6GAL1:70, ST6GAL2:85; Sample ID5, renal clear cells, ST3GAL1:105, ST3GAL2:155, ST6GAL1:195, ST6GAL2:175; Sample ID6, normal kidney cells, ST3GAL1:55, ST3GAL2:75, ST6GAL1:75, ST6GAL2:88; Samples ID1, ID2, ID3, and ID4 were used as cell training sets, and ID5 and ID6 were used as cell validation sets.
[0038] S2. Calculate fold change and t-test: ST3GAL1: fold change = 1.91, p value = 0.0021; ST3GAL2: fold change = 2.09, p value = 0.0018; ST6GAL1: fold change = 2.50, p value = 0.0009; ST6GAL2: fold change = 2.00, p value = 0.0023; The p-value threshold was set to 0.05 and the fold change threshold was set to 1.5. All genes met the conditions, so the candidate gene set contained all four genes.
[0039] S3. Determination of the best gene signature: All possible gene combinations (15 in total) were generated. It was assumed that after calculation by the random forest algorithm, the best gene signature was obtained: ST3GAL2, ST6GAL1.
[0040] S4. Constructing a prediction model based on the best gene signature: The prediction model was constructed using the expression data of ST3GAL2 and ST6GAL1; Make predictions on the validation set: Sample ID5 prediction result: kidney clear cell; Sample ID6 prediction result: normal kidney cells; Verify the prediction results: ST3GAL2 validation fold change: 2.07 (meets threshold); ST6GAL1 validation fold change: 2.60 (met the threshold).
[0041] Final Model: The prediction model based on the expression data of ST3GAL2 and ST6GAL1 can be used to predict new cell sample types.
[0042] In actual applications, the number of samples will be larger, the number of genes will be larger, and the calculation process will be more complicated, but the basic principles and steps are the same.
[0043] The second purpose of this embodiment is to provide a method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes in cell gene analysis.
[0044] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. A method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes, characterized in that: The steps are as follows: S1. Collect cell samples, wherein the cell samples include cell sample ID, sample type and sialyltransferase gene expression data; divide the cell samples into a cell training set and a cell validation set, and select renal clear cell samples and normal renal cell samples in the cell training set as the total gene set; S2. Based on the sialyltransferase gene expression data of renal clear cell samples and the sialyltransferase gene expression data of normal renal cells in the total gene set, the gene differences between the two were compared using fold change and t-test methods, and a candidate gene set was generated; S3. Based on the candidate gene set, a feature selection model is constructed using the random forest algorithm. Based on the constructed feature selection model, the most discriminative gene combination is selected as the best gene signature; S4. Select the corresponding sialyltransferase gene expression data in the total gene set according to the best gene signature to generate a gene signature data set; and construct a prediction model using a random forest algorithm based on the gene signature data set; In addition, renal clear cell samples and normal renal cell samples were selected from the sample types in the cell validation set as the validation gene set, and the sample type of the validation gene set was predicted using the constructed prediction model; based on the predicted sample type, the prediction model was verified using the fold change.
2. The method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes according to claim 1, characterized in that: The method uses fold change and t-test method to compare the gene differences between the two and generate a candidate gene set, which specifically includes: For the expression value corresponding to each gene type of sialyltransferase gene expression data in the total gene set, calculate its mean value and standard deviation in renal clear cell samples and normal renal cell samples; and calculate the ratio of the mean value of each gene type in renal clear cell samples and normal renal cell samples to generate the fold change results of each gene type in renal clear cell samples and normal renal cell samples; The t statistic was calculated based on the mean and standard deviation of each gene type in renal clear cell samples and normal renal cell samples, and the formula is as follows: ,in and Respectively represent the average values of gene types in renal clear cell samples and normal renal cell samples, and The standard deviation of the gene types in the renal clear cell samples and normal renal cell samples, respectively, and The number of samples with the respective genotypes in samples representing renal clear cells and samples representing normal renal cells; Calculating degrees of freedom , the formula is According to the t statistic and degrees of freedom, the t distribution table is used to obtain the Value, where The value is a probability value used in statistics to quantify the degree of difference between the observed data and the null hypothesis; set up The threshold of the value and the threshold of the fold change are set; the screening The gene types with the best values and fold change thresholds were selected as candidate gene sets.
3. The method for labeling renal clear cells using the sialyltransferase gene signature for non-disease diagnosis purposes according to claim 2, characterized in that: The construction of the feature selection model using the random forest algorithm specifically includes: The candidate gene set was organized into a table, where the rows represented different cell sample IDs and the columns included sialyltransferase gene type, expression value, and sample type; Using the combination principle, the candidate gene sets are used to generate all combination sets; Select a gene combination from the combination set in the order of the list as the current gene combination, extract the corresponding expression value and sample type in the table according to the current gene combination, and use it as the training set of the feature selection random forest algorithm; Samples are randomly extracted with replacement from a training set of a feature selection random forest algorithm, and each node of a decision tree in the feature selection random forest is constructed by randomly selecting a features as candidates from the current gene combination in the training set of the feature selection random forest algorithm, where a is the square root of the total number of gene types in the current gene combination; recursively construct until a stopping condition is reached, and the stopping condition includes but is not limited to reaching a preset maximum tree depth and the number of samples in the node is less than a preset minimum number.
4. The method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes according to claim 3, characterized in that: The method of selecting the most discriminative gene combination as the best gene signature based on the constructed feature selection model specifically includes: A decision tree is constructed for each gene combination in the combination set to form a feature selection model. The average importance score of each gene combination is calculated based on the decision tree corresponding to the gene. The average importance is based on the degree to which the gene reduces impurity in the tree. For each gene in the combination, the average importance score in all decision trees is calculated. The average importance scores of all genes in the combination are added and averaged to obtain the average importance score of the combination. All gene combinations were ranked in descending order according to the average importance score of each combination, and the combination with the highest score was selected as the best gene signature that could effectively distinguish sialyltransferases from renal clear cells.
5. The method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes according to claim 1, characterized in that: The construction of the prediction model using the random forest algorithm specifically includes: According to the gene combination in the best gene signature, the expression data and sample type of the sialyltransferase gene corresponding to the gene combination are extracted from the total gene set and used as the gene signature data set; According to the gene signature dataset, the prediction model is constructed using the random forest algorithm. Each prediction decision tree is randomly sampled with replacement. At each branch node, a subset of the gene signature dataset is randomly selected for splitting, and the process is repeated until the prediction model converges.
6. The method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes according to claim 5, characterized in that: The method of using the constructed prediction model to predict the sample type of the validation gene set specifically includes: For the validation gene set, the gene expression values corresponding to the best gene signature are extracted and input into each prediction decision tree in the prediction model. Each prediction decision tree gives a prediction result. The prediction results of all prediction decision trees are summarized, and the majority voting method is used to determine the final predicted sample type.
7. The method for labeling renal clear cells using a sialyltransferase gene signature for non-disease diagnosis purposes according to claim 6, characterized in that: The predicted validation gene set is validated using fold changes, specifically including: For the validation gene set and the predicted sample type, the average value of each gene in the validation gene set in the predicted sample type is calculated, and the average value is compared with the average value of the actual sample type to generate the validation fold change; The validation fold change is compared with the set fold change threshold. If the threshold requirement is met, its prediction model is output as the final model, which will predict the sample type of the new cell based on the gene type in the best gene signature.
8. Application of the sialyltransferase gene signature for non-disease diagnosis purposes according to any one of claims 1 to 7 for labeling of renal clear cells in cell gene analysis.