A method for analyzing and processing gene microarray data

Abstract

The application discloses a kind of gene microarray data analysis processing methods, it is related to biomedical information processing technical field, comprising: using microarray technology, obtain gene expression data;Multiple low-dimensional feature selection tasks are constructed;Optimal feature subset is obtained by multi-task pseudo-affine transformation algorithm;According to optimal feature subset, neural network model is trained;Analysis processing is carried out to the gene to be predicted and whether it is predicted to be ill.The gene microarray data analysis processing method of the application is analyzed to gene microarray data characteristics, and gene microarray data is helpful to dig disease characteristic gene information, and has a key role to early detection, clinical treatment and disease prevention of disease.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biomedical information processing, and particularly relates to a gene microarray data analysis processing method. BACKGROUND

[0002] Gene chip technology is an efficient and large-scale gene data acquisition technology, which can simultaneously study the relationship between tens of thousands of gene expressions and diseases. The gene chip technology can be used for gene mutation diagnosis, new drug research and development, disease diagnosis and individualized medical treatment and many other applications. Through data analysis, the nature of cancer can be revealed at the gene level. In biomedicine, the characteristics of gene expression data (also known as microarray data) are high dimensionality, small sample size, and extremely rare genes that are truly related to diseases. Due to the high dimensionality of the data, the subsequent classification is inefficient, and too many genes will increase noise and affect the classification effect.

[0003] In fact, only a small number of key genes affect the classification of all samples. Therefore, the key problem is to use an efficient data analysis algorithm and find the effective genes in the gene expression profile data through calculation. Feature selection is an important data preprocessing step. The main idea of feature selection is to select the most valuable feature subset by deleting irrelevant and redundant features from the feature space of the original data set, so as to improve the prediction accuracy, robustness and interpretability of the model. It can be divided into four steps: generating a feature subset, evaluating a feature subset, setting a stopping criterion and judging whether the stopping is sufficient, and verifying the final result. Assuming that there are n features, each feature can be selected or not selected, so there are 2n cases of feature subsets. When n is very large, it is obviously not feasible to obtain the best feature subset by exhaustive selection due to time complexity. Therefore, it is an important problem to be considered and solved to quickly and effectively find the best features from the feature space.

[0004] In summary, it is urgent to study a gene microarray data analysis processing method. SUMMARY

[0005] To solve the above technical problems, the present application discloses a gene microarray data analysis processing method, which analyzes the characteristics of gene microarray data, helps to mine disease characteristic gene information from gene microarray data, and plays a key role in early detection, clinical treatment and disease prevention of diseases.

[0006] To achieve the above object, the technical scheme adopted by the present application is as follows:

[0007] The present application provides a gene microarray data analysis processing method in the first aspect, which comprises the following steps:

[0008] Step S1. Obtain gene expression data by using microarray technology;

[0009] Step S2. Constructing multiple low-dimensional feature selection tasks;

[0010] Step S3. Obtaining the optimal feature subset by a multi-task pseudo-affine transformation algorithm;

[0011] Step S4. Training a neural network model according to the optimal feature subset;

[0012] Step S5. Analyzing and processing the to-be-predicted gene and predicting whether the gene is diseased.

[0013] Optionally, in step S1, the step of obtaining gene expression data by using a microarray technology specifically comprises:

[0014] Step S11. Preparing a gene chip: fixing probes on a carrier to form a DNA microarray;

[0015] Fixing oligonucleotides, polynucleotides or cDNA fragments called probes on a carrier such as a glass slide, a quartz chip or a nylon membrane in a specific arrangement to form a DNA microarray;

[0016] Step S12. Preparing fluorescently labeled probes: before labeling the sample into probes, purification, reverse transcription or amplification is needed. During the PCR (polymerase chain reaction) amplification process, the sample needs to be labeled at the same time;

[0017] Step S13. Hybridization of labeled probes and chip: select appropriate hybridization conditions according to the type, length of the probe and the application of the chip, hybridize the prepared fluorescent probes with the chip, wash off the unbound probes after the appropriate elution step, and excite the sample on the chip with a laser to emit fluorescence;

[0018] Step S14. Scanning the hybridization map: excite the hybridized chip with a laser of a specific wavelength, and the probes on the chip will emit fluorescence of different wavelengths. Then, the fluorescence intensity of the probes is detected by a laser confocal fluorescence detection system. The fluorescence signal of strictly paired hybridized molecules is strong, and the fluorescence of incompletely hybridized molecules is weak;

[0019] Step S15. Data extraction: through special software processing and analysis, gene expression data can be extracted from the hybridization image, i.e., the original hybridization image is converted into gene expression profile data.

[0020] Optionally, in step S2, the step of constructing multiple low-dimensional feature selection tasks specifically comprises:

[0021] Step S21. Analyzing the weight of each gene by a ReliefF and Minimum Redundancy Maximum Relevance (MRMR) filter type feature analysis method;

[0022] Step S22. Constructing two feature selection tasks based on the ReliefF algorithm.

[0023] Step S23. Constructing two feature selection tasks based on the MRMR algorithm.

[0024] Optionally, in step S22, the step of constructing two feature selection tasks based on the ReliefF algorithm specifically includes:

[0025] Step S221. Obtaining the weight of each gene through the ReliefF algorithm.

[0026] Step S222. Generating a random number and comparing the weight threshold of ReliefF with the size of the random number.

[0027] Step S223. Repeating step S222 to determine whether each gene is selected, and constructing the first low-dimensional feature selection task with all the selected genes.

[0028] Step S224. Repeating step S222 to determine whether each gene is selected, and constructing the second low-dimensional feature selection task with all the selected genes.

[0029] Optionally, in step S23, the step of constructing two feature selection tasks based on the MRMR algorithm specifically includes:

[0030] Step S231. Obtaining the weight of each gene through the MRMR algorithm.

[0031] Step S232. Generating a random number and comparing the weight threshold of MRMR with the size of the random number.

[0032] Step S233. Repeating step S232 to determine whether each gene is selected, and constructing the third low-dimensional feature selection task with all the selected genes.

[0033] Step S234. Repeating step S232 to determine whether each gene is selected, and constructing the fourth low-dimensional feature selection task with all the selected genes.

[0034] Optionally, in step S3, the step of obtaining the optimal feature subset through the multi-task pseudo-affine transformation algorithm includes:

[0035] Step S31. Setting the population size n and the maximum number of iterations T; randomly initializing the population X and using formula (1) as the objective function

[0036] f(x) = 0.99 * accuracy + 0.01 * |S| / |C| (1)

[0037] Wherein, accuracy represents the prediction accuracy obtained by using the classification algorithm, |S| represents the number of selected genes, and |C| represents the number of all genes.

[0038] Step S32. Initialize the selection matrix M with a lower triangular matrix tmp , randomly shuffle the rows and columns of M tmp to obtain the final selection matrix M.

[0039] Step S33. Calculate the inverse matrix of the selection matrix is obtained by binary inverse operation on the matrix elements of M, and the binary inverse operation means that the element 0 in the matrix M will be inversely transformed into 1, and the element 1 will be inversely transformed into 0.

[0040] Step S34. Calculate the mutation matrix B, generate a random number, and compare it with a specified threshold value. If the random number is less than the threshold value, inter-task communication is performed, otherwise inter-task communication is performed.

[0041] Step S35. Convert each element in the matrix B to a binary form by a sigmoid transfer function, as shown in equation (4),

[0042]

[0043] Wherein, rand represents a random number in [0, 1];

[0044] Step S36. Get the new solution position X by equation (5),

[0045]

[0046] Wherein, represents the dot product of the matrix, which is the same as the “.*” operation in Matlab;

[0047] Step S37. Determine whether the new solution position is better than the original position. If yes, replace the original position.

[0048] Step S38. Repeat steps S32-S37 until the program meets the conditions to exit, and obtain the optimal solution of each subtask.

[0049] Step S39. Calculate the optimal solution of each subtask on the overall task. The solution with the best fitness is the final solution of the multi-task algorithm, and the features selected by the final solution constitute the optimal feature subset.

[0050] Optionally, in step S34, the calculation of the mutation matrix B includes:

[0051] Step S341. There are three ways to calculate the mutation matrix B for inter-task communication:

[0052] The first: B = X gbest +F*(X r1 -X r2 )

[0053] The second: B = X r3 +F*(X r1 -X r2 )

[0054] The third: B = X + F*(X r1 -X r2 )

[0055] Wherein, X gbest represents the current optimal solution of the task, F represents the size factor, usually taking the value of [0, 2]; X r1 , X r2 and X r3 are optional solutions on the task; X represents the current solution;

[0056] Step S342. There are three ways to calculate the variation matrix B of inter-task communication: the first:

[0057] The second:

[0058] The third:

[0059] Wherein, represents an optional solution from task t, represents an optional solution from task g.

[0060] Optionally, in step S4, the step of training the neural network model according to the optimal feature subset comprises:

[0061] Step S41. According to the optimal feature subset, feature selection is performed on all sample data;

[0062] Step S42. The processed data is divided into a training set and a test set;

[0063] Step S43. The training set data is input into the neural network;

[0064] Step S44. Adjust the network parameters by gradient descent;

[0065] Step S45. Test the model effect by 10-fold cross-validation to obtain the optimal network model.

[0066] Optionally, in step S5, the step of analyzing and processing the to-be-predicted gene and predicting whether it is diseased comprises:

[0067] Step S51. Obtain gene expression data of the analyzed genes through gene chip technology;

[0068] Step S52. Perform feature selection on the genes to be predicted according to the optimal feature subset of step S3;

[0069] Step S53. Input the data obtained in step S52 into the neural network model of step S4;

[0070] Step S54. Obtain the disease prediction result through the neural network model.

[0071] The second aspect of the present application also provides a computer readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the gene microarray data analysis processing method of the second aspect of the present application.

[0072] The gene microarray data analysis processing method based on the multi-task pseudo affine transformation algorithm of the present application digitizes gene data through gene chip technology, processes feature selection of the digitized gene data through multi-task evolutionary computation, decomposes the high-dimensional gene selection task into multiple low-dimensional subtasks through the ReliefF and MRMR algorithms, which can reduce the task complexity and improve the processing speed, and facilitates parallel processing of multiple subtasks. Based on the multi-task pseudo affine transformation algorithm, information exchange between subtasks can be considered while processing respective subtasks, which can help the algorithm to jump out of the local optimal solution, improve the classification accuracy, and speed up the prediction process.

[0073] The method of the present application is used for analyzing the relationship between genes and diseases, identifying and predicting the occurrence of diseases. Through feature analysis of gene microarray data, gene microarray data is helpful for mining disease characteristic gene information, and plays a key role in early detection, clinical treatment and disease prevention of diseases. BRIEF DESCRIPTION OF DRAWINGS

[0074] Figure 1 A gene microarray data analysis processing method flowchart of the present application;

[0075] Figure 2 A method flowchart for obtaining an optimal feature subset in a gene microarray data analysis processing method of the present application;

[0076] Figure 3 A neural network model schematic diagram of a gene microarray data analysis processing method of the present application;

[0077] Figure 4 A multi-task pseudo affine transformation algorithm overall flowchart schematic diagram of a gene microarray data analysis processing method of the present application;

[0078] Figure 5A schematic diagram of information exchange between tasks in a multi-task pseudo-affine transformation algorithm for a gene microarray data analysis processing method of the present application;

[0079] Figure 6 A flow chart of processing of a gene to be predicted in a gene microarray data analysis processing method of the present application. DETAILED DESCRIPTION

[0080] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0081] Embodiment 1

[0082] A gene microarray data analysis processing method, as shown in Figure 1 includes the following steps:

[0083] Step S1. Obtain gene expression data by using microarray technology;

[0084] Specifically includes:

[0085] Step S11. Prepare a gene chip: fix oligonucleotides, skin nucleotides or cDNA fragments called probes on a carrier such as a glass slide, a quartz chip or a nylon membrane in a specific arrangement to form a DNA microarray;

[0086] Step S12. Prepare a fluorescently labeled probe: before the sample is labeled into a probe, it needs to be purified, reverse transcribed or amplified. During the PCR (polymerase chain reaction) amplification process, the sample needs to be labeled at the same time;

[0087] Step S13. Hybridize the labeled probe to the chip: select appropriate hybridization conditions according to the type, length of the probe and the application of the chip, etc. Hybridize the prepared fluorescent probe to the chip, wash off the unbound probe after the appropriate elution step, and excite the sample on the chip with a laser to emit fluorescence;

[0088] Step S14. Scan the hybridization map: excite the hybridized chip with a laser of a specific wavelength, and the probe on the chip will emit fluorescence of different wavelengths. Then, detect the fluorescence intensity of the probe by using a laser confocal fluorescence detection system. The fluorescence signal of strictly paired hybridized molecules is strong, and the fluorescence of incompletely hybridized molecules is weak;

[0089] Step S15. Data extraction: through processing and analysis by using special software, gene expression data can be extracted from the hybridization image, i.e., the original hybridization image is converted into gene expression profile data;

[0090] Step S2. Constructing multiple low-dimension feature selection tasks;

[0091] Specifically comprising:

[0092] Step S21. Analyzing each gene weight by ReliefF and MRMR filtering feature analysis method;

[0093] Step S22. Constructing 2 feature selection tasks based on ReliefF algorithm;

[0094] Optionally, the step specifically comprises:

[0095] Step S221. Obtaining each gene weight by ReliefF algorithm;

[0096] Step S222. Generating a random number, comparing the weight threshold value of ReliefF and the size of the random number;

[0097] Step S223. Repeating step S222 to determine whether each gene is selected, and all selected genes constitute the first low-dimension feature selection task;

[0098] Step S224. Repeating step S222 to determine whether each gene is selected, and all selected genes constitute the second low-dimension feature selection task;

[0099] Step S23. Constructing 2 feature selection tasks based on MRMR algorithm;

[0100] Optionally, the step specifically comprises:

[0101] Step S231. Obtaining each gene weight by MRMR algorithm;

[0102] Step S232. Generating a random number, comparing the weight threshold value of MRMR and the size of the random number;

[0103] Step S233. Repeating step S232 to determine whether each gene is selected, and all selected genes constitute the third low-dimension feature selection task;

[0104] Step S234. Repeating step S232 to determine whether each gene is selected, and all selected genes constitute the fourth low-dimension feature selection task;

[0105] Step S3. Obtaining the optimal feature subset by multi-task pseudo-affine transformation algorithm, as shown in Figure 2 and 4 ;

[0106] Specifically comprising:

[0107] Step S31. Set population number n, maximum iteration number T; randomly initialize population X, and use formula (1) as objective function

[0108] f(x) = 0.99 * accuracy + 0.01 * |S| / |C (1)

[0109] Wherein, accuracy represents the prediction accuracy obtained using the classification algorithm, |S| represents the number of selected genes, and |C| represents the number of all genes.

[0110] Step S32. Initialize selection matrix M with lower triangular matrix tmp , randomly shuffle the rows and columns of M tmp to obtain the final selection matrix M; as shown in formula (2):

[0111]

[0112] Step S33. Calculate the inverse matrix of the selection matrix is obtained by binary inverse operation on the matrix elements of M, and the binary inverse operation means that the element 0 in the matrix M will be inversely transformed into 1, and the element 1 will be inversely transformed into 0; as shown in formula (3):

[0113]

[0114] Step S34. Calculate the mutation matrix B, generate a random number, and compare it with the specified threshold value. If the random number is less than the threshold value, the intra-task communication is performed, otherwise the inter-task communication is performed.

[0115] Optionally, the step specifically includes:

[0116] Step S341. The calculation method of the mutation matrix B of the intra-task communication includes three kinds, and one of them is selected:

[0117] The first kind: B = X gbest +F*(X r1 -X r2 )

[0118] The second kind: B = X r3 +F*(X r1 -X r2 )

[0119] The third kind: B = X+F*(X r1 -X r2 )

[0120] Wherein, X gbest represents the current optimal solution of this task, and F represents the size factor, which is usually taken as [0, 2]; X r1 , X r2 and Xr3 Xopt is the optional solution of task t; X denotes the current solution;

[0121] Step S342. There are three ways to calculate the variation matrix B of inter-task communication, one of which is optional:

[0122] The first way:

[0123] The second way:

[0124] The third way:

[0125] wherein, Xopt is the optional solution of task t; X denotes the current solution; Xopt is the optional solution of task g.

[0126] As shown in Figure 5 , that is, the calculation of There are 10 genes in total, task t is performed on genes 1, 3, 4, 7, and 9 for feature selection, and task g is performed on genes 2, 3, 4, 5, 7, and 8 for feature selection, The solution is [1, 1, 0, 1, 0], The solution is [1, 1, 0, 1, 1, 0]. Figure 5 First, transform and to all genes, then perform subtraction operation, and then convert the difference to t task.

[0127] Step S35. Convert each element in matrix B to binary form by sigmoid transfer function, as shown in equation (4),

[0128]

[0129] wherein, rand represents a random number of [0, 1];

[0130] Step S36. Obtain the position X of the new solution by equation (5),

[0131]

[0132] wherein, denotes the dot product of the matrix, which is the same as the “.*” operation in Matlab;

[0133] Step S37. Determine whether the new solution position is better than the original position, if so, replace the original position;

[0134] Step S38. Repeat steps S32-S37 until the program meets the conditions to exit, and obtain the optimal solution of each sub-task;

[0135] Step S39. Calculate the optimal solution for each subtask on the overall task. The solution with the best fitness is the final solution of the multi-task algorithm. The features selected by the final solution form the optimal feature subset.

[0136] Step S4. Train the neural network model based on the optimal feature subset;

[0137] Specifically, it includes:

[0138] Step S41. Based on the optimal feature subset, perform feature selection on all sample data;

[0139] Step S42. Divide the processed data into a training set and a test set;

[0140] Step S43. Input the training set data into the neural network;

[0141] Step S44. Adjust network parameters using gradient descent;

[0142] Step S45. Use 10-fold cross-validation to test the model's performance and obtain the optimal network model;

[0143] like Figure 3 As shown, the specific structure of the neural network model includes:

[0144] The input layer has the same number of nodes as the number of features in the optimal feature subset.

[0145] Hidden layer 1, with 10 hidden nodes;

[0146] Hidden layer 2, with 20 hidden nodes;

[0147] Hidden layer 3, with 20 hidden nodes;

[0148] Hidden layer 4, with 10 hidden nodes;

[0149] The output layer has 2 nodes, representing: diseased and non-diseased.

[0150] Step S5. Analyze and process the gene to be predicted and predict whether it will cause disease;

[0151] like Figure 6 As shown, it specifically includes:

[0152] Step S51. Obtain gene expression data of the analyzed gene using gene chip technology;

[0153] Step S52. Select features for the gene to be predicted based on the optimal feature subset obtained in step S3;

[0154] Step S53. Input the data obtained in step S52 into the neural network model in step S4;

[0155] Step S54. Obtain disease prediction result through neural network model.

[0156] Embodiment 2

[0157] A computer readable storage medium, which stores a program, the program is executed by a processor to realize the steps in the gene microarray data analysis processing method described in embodiment 1, comprising:

[0158] Step S1. Obtain gene expression data by using microarray technology;

[0159] Step S2. Construct a plurality of low-dimensional feature selection tasks;

[0160] Step S3. Obtain optimal feature subset through multi-task pseudo-affine transformation algorithm;

[0161] Step S4. Train neural network model according to optimal feature subset;

[0162] Step S5. Analyze and process the genes to be predicted and predict whether to be sick.

[0163] Of course, the above description is not a limitation of the present application, and the present application is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present application should also be within the scope of the present application.

Claims

1. A method for analyzing and processing gene microarray data, characterized in that, Includes the following steps: Step S1. Obtain gene expression data using microarray technology; Step S2. Construct multiple low-dimensional feature selection tasks; Step S3. Obtain the optimal feature subset using a multi-task pseudo-affine transformation algorithm; Step S4. Train the neural network model based on the optimal feature subset; Step S5. Analyze and process the gene to be predicted and predict whether it will cause disease; Step S3, the step of obtaining the optimal feature subset through the multi-task pseudo-affine transformation algorithm, includes: Step S31. Set the population size n and the maximum number of iterations T; randomly initialize the population. And use the following formula as the objective function, ； Where accuracy represents the prediction accuracy obtained using the classification algorithm, |S| represents the number of genes selected, and |C| represents the total number of genes. Step S32. Initialize the selection matrix using a lower triangular matrix. Randomly shuffle By dividing the rows and columns, we obtain the final selection matrix M; Step S33. Calculate the inverse matrix of the selection matrix. , It is obtained by performing binary inverse operations on the matrix elements of M. The binary inverse operation means that element 0 in matrix M will be transformed into 1, and element 1 will be transformed into 0. Step S34. Calculate the mutation matrix B, generate random numbers, and compare them with a specified threshold. If the random number is less than the threshold, then communicate within the subtask; otherwise, communicate between subtasks. Step S35. Convert each element in matrix B into binary form using the sigmoid transfer function, as shown in the following equation. ； Where rand represents a random number in the range [0,1]; Step S36. Obtain the position of the new solution using the following formula. , ； in, This represents the dot product of matrices, and is the same as the ".*" operation in Matlab. Step S37. Determine whether the new solution position is better than the original position. If it is better, replace the original position. Step S38. Repeat steps S32-S37 until the program exits under the condition that it meets the condition, and obtain the optimal solution for each subtask; Step S39. Calculate the optimal solution for each subtask on the overall task. The solution with the best fitness is the final solution of the multi-task algorithm. The features selected by the final solution form the optimal feature subset. In step S34, the mutation matrix B is calculated, including: Step S341. The calculation method for the mutation matrix B of intra-task communication includes: The first type: ; The second type: ; The third type: ; in, This represents the current optimal solution for this task. This represents the size factor, which typically takes values ​​of [0, 2]. , and The solution chosen for this task; Indicates the current solution; Step S342. The calculation method for the mutation matrix B of inter-task communication includes: The first type: ; The second type: ; The third type: ; in, This represents a solution chosen from task t. This represents a solution chosen from task g.

2. The gene microarray data analysis and processing method as described in claim 1, characterized in that, Step S1, which involves obtaining gene expression data using microarray technology, specifically includes: Step S11. Gene chip preparation: Immobilize probes on a vector to form a DNA microarray; Step S12. Prepare fluorescently labeled probes; Step S13. Hybridize the labeled probes with the chip; Step S14. Scan the hybridization map; Step S15. Data extraction: Extract gene expression data from the hybridization image and convert the original hybridization image into gene expression profile data.

3. The gene microarray data analysis and processing method as described in claim 1, characterized in that, Step S2, which involves constructing multiple low-dimensional feature selection tasks, specifically includes: Step S21. Analyze the weight of each gene using the ReliefF and minimum redundancy maximum correlation filtering feature analysis methods; Step S22. Construct two feature selection tasks based on the ReliefF algorithm; Step S23. Construct two feature selection tasks based on the minimum redundancy maximum correlation algorithm.

4. The gene microarray data analysis and processing method as described in claim 3, characterized in that, Step S22, which involves constructing two feature selection tasks based on the ReliefF algorithm, specifically includes: Step S221. Obtain the weight of each gene using the ReliefF algorithm; Step S222. Generate a random number and compare the weight threshold of ReliefF with the size of the random number; Step S223. Repeat step S222 to determine whether each gene is selected, and make all the selected genes into the first low-dimensional feature selection task. Step S224. Repeat step S222 to determine whether each gene is selected, and make all the selected genes into the second low-dimensional feature selection task.

5. The gene microarray data analysis and processing method as described in claim 3, characterized in that, Step S23, which involves constructing two feature selection tasks based on the minimum redundancy maximum relevance algorithm, specifically includes: Step S231. Obtain the weight of each gene using the minimum redundancy maximum correlation algorithm; Step S232. Generate a random number and compare the weight threshold of minimum redundancy and maximum relevance with the size of the random number; Step S233. Repeat step S232 to determine whether each gene is selected, and make all the selected genes into the third low-dimensional feature selection task. Step S234. Repeat step S232 to determine whether each gene is selected, and make all the selected genes into the fourth low-dimensional feature selection task.

6. The gene microarray data analysis and processing method as described in claim 1, characterized in that, Step S4, the step of training the neural network model based on the optimal feature subset, includes: Step S41. Based on the optimal feature subset, perform feature selection on all sample data; Step S42. Divide the processed data into a training set and a test set; Step S43. Input the training set data into the neural network; Step S44. Adjust network parameters using gradient descent; Step S45. Use 10-fold cross-validation to test the model's performance and obtain the optimal network model.

7. The gene microarray data analysis and processing method as described in claim 1, characterized in that, The steps for analyzing and processing the gene to be predicted and predicting whether a disease will occur include: Step S51. Obtain the gene expression data for analysis using gene chip technology; Step S52. Select features for the gene to be predicted based on the optimal feature subset obtained in step S3; Step S53. Input the data obtained in step S52 into the neural network model in step S4; Step S54. Obtain disease prediction results through a neural network model.

8. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by a processor, the program implements the steps in the gene microarray data analysis and processing method as described in any one of claims 1 to 7.

Images