Method for diagnosing refractory depression by using blood gene expression combination

Abstract

The invention provides a method for diagnosing refractory depression by using a blood gene expression combination, and belongs to the fields of biotechnology and medicine. According to the method, the RNA-seq technology and the machine learning algorithm are combined, and the classification model based on the blood transcriptome expression matrix is constructed, so that rapid and accurate diagnosis of the TRD patient is realized. Specifically, firstly, blood gene expression matrixes of a TRD patient and a normal person are obtained, a lima packet is utilized to remove a data batch effect and perform differential gene expression analysis, and differential genes are selected as features to be input into classification models such as a support vector machine (SVM), a random forest (RF) and a one-dimensional convolutional neural network (1D CNN) for training. The trained model can output a TRD diagnosis result according to a blood gene expression matrix of a patient. The method disclosed by the invention has the advantages of low cost, high efficiency and high accuracy, and has a wide application prospect in clinical diagnosis of TRD.

Description

Technical Field

[0001] The present invention belongs to the field of biotechnology and medicine, and relates to a method for diagnosing TRD patients based on machine learning combined with blood transcriptomics, and more specifically to a TRD diagnosis method using a combination of blood gene expression. Background Art

[0002] Treatment-resistant depression (TRD) represents a great complexity in both its research and treatment. The definition itself has 150 different attempts to formulate it. Due to the lack of diagnostic and therapeutic approaches with predictive value, the criteria proposed by the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) - failure to respond to two or more antidepressants at appropriate doses and durations and treatment compliance - still hold reference status. It is indisputable that TRD is a topic of high social relevance, as an estimated 30-55% of patients with depression do not achieve full remission and are at high risk of suicide.

[0003] RNA-seq stands for RNA sequencing, which is a high-throughput gene expression analysis technology that allows researchers to conduct comprehensive quantitative analysis of RNA molecules in cells at the molecular level. RNA-seq technology can reveal transcriptome characteristics under specific biological states, including gene expression levels, transcript isoforms, alternative splicing events, and the expression of non-coding RNA. It can be applied in many fields, such as: disease mechanism research: revealing the molecular mechanism of disease occurrence by comparing the differences in gene expression between healthy and diseased states; drug development: identifying molecular targets of drug action and evaluating the effects of drugs on gene expression; personalized medicine: analyzing individual transcriptome data to provide a basis for personalized treatment plans; cell heterogeneity research: single-cell RNA-seq technology can reveal the heterogeneity within the cell population and identify different cell subpopulations.

[0004] Machine learning refers to the process of computers simulating human learning behaviors through large amounts of data training and analysis to acquire new knowledge and skills. It has been successfully applied to many fields, including computer vision, speech and handwriting recognition, and autonomous driving. With the development of computers, machine learning algorithms can process and analyze large amounts of medical data, including electronic health records, medical images, genomic data, etc., to identify disease patterns and predict the development of diseases. This has led to the rapid development of machine learning in disease prediction, which is expected to further improve the efficiency of disease prevention, diagnosis, and treatment in the future.

[0005] Since there are many different ways to diagnose refractory depression, although the standard reference is now generally the lack of response to two or more antidepressants at appropriate doses and durations as well as treatment compliance, this still requires a lot of time and energy, so there is an urgent need for a method to quickly and accurately identify TRD patients. Summary of the invention

[0006] The present invention provides a TRD diagnosis method using a combination of blood gene expression. Specifically, the present invention aims to construct a diagnostic method that can quickly and accurately determine TRD patients by combining RNA-seq technology and machine learning algorithms.

[0007] The technical solution of the present invention comprises the following steps:

[0008] A method for diagnosing treatment-resistant depression using a blood gene expression panel, comprising:

[0009] Obtain a blood differential gene expression matrix of the patient to be tested; wherein the blood differential genes are differential genes obtained by screening the blood transcriptomes of TRD patients and normal people, and further screened based on the weights calculated by the machine model to obtain differential genes with weights greater than those of important inflammatory factors of TRD; important inflammatory factors of TRD are one or more of IL-6, IL-1β and TNF-α;

[0010] Inputting the blood differential gene expression matrix into a trained machine classification model, and diagnosing whether the patient to be tested suffers from refractory depression according to the diagnosis result output by the classification model;

[0011] The trained machine classification model is based on a training data set, takes a blood differential gene expression matrix as input, takes a predicted diagnosis result as output, and is trained by minimizing the error between the output of the classification model and the TRD label as the goal.

[0012] Furthermore, the differentially expressed genes obtained by screening the blood transcriptomes of TRD patients and normal subjects were obtained by screening using the following method:

[0013] Collect blood transcriptome data from TRD and normal subjects;

[0014] Differential gene analysis was performed with TRD as the experimental group and normal subjects as the control group, and the first group of genes obtained by screening were the differential genes.

[0015] Furthermore, the screening threshold was P-value≤0.05,|log2FC|≥log2(1.2).

[0016] Furthermore, it also includes further performing KEGG and GO-BP enrichment analysis on the first group of genes obtained by screening, and the second group of genes obtained by screening are differential genes.

[0017] Furthermore, the machine model is a random forest or a support vector machine; wherein, in the support vector machine, the blood differential gene expression matrix composed of the first group of genes or the second group of genes obtained by screening is used as input for training, and the model characteristic coefficient is used to obtain the weight corresponding to each gene feature; in the random forest, the blood differential gene expression matrix composed of the first group of genes or the second group of genes obtained by screening is used as input for training, and the feature importance is used to obtain the weight corresponding to each gene feature.

[0018] Furthermore, each sample of the training data set includes a blood differential gene expression matrix and a corresponding TRD label.

[0019] Furthermore, the machine classification model includes a classification model constructed based on a support vector machine, a classification model constructed based on a random forest and / or a one-dimensional convolutional neural network.

[0020] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for diagnosing refractory depression using a combination of blood gene expressions is implemented.

[0021] A storage medium containing computer executable instructions, which, when executed by a computer processor, implement a method for diagnosing refractory depression using a blood gene expression combination.

[0022] A computer program product comprises a computer program / instruction, which, when executed by a processor, implements the steps of a method for diagnosing refractory depression using a blood gene expression combination.

[0023] The present invention provides a TRD diagnostic method with low cost, high efficiency and high accuracy. The method can effectively utilize blood gene expression data, provides convenience for the diagnosis of a large number of patients, and has broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A general construction and training flow chart of the machine classification model of the present invention;

[0025] Figure 2 This is a graph showing the weight of each differentially expressed gene in a training set calculated by RF in an optional embodiment of the present invention;

[0026] Figure 3 This is a graph showing the weight of each differential gene in a training set calculated by SVM in an optional embodiment of the present invention;

[0027] Figure 4The AUC best score result graph of the TRD diagnostic classification prediction model obtained by the optimal training set obtained by screening the differential gene expression matrix based on logFC or weight;

[0028] Figure 5 It is the classification index score of the TRD diagnostic classification prediction model obtained by the optimal training set obtained by screening the differential gene expression matrix based on logFC or weight. DETAILED DESCRIPTION

[0029] The present invention provides a method for diagnosing refractory depression using a blood gene expression panel, comprising:

[0030] Obtain a blood differential gene expression matrix of the patient to be tested; wherein the blood differential genes are differential genes obtained by screening the blood transcriptomes of TRD patients and normal people, and further screened based on the weights calculated by the machine model to obtain differential genes with weights greater than those of important inflammatory factors of TRD; important inflammatory factors of TRD are one or more of IL-6, IL-1β and TNF-α;

[0031] The blood differential gene expression matrix is ​​input into a trained machine classification model, and whether the patient to be tested suffers from refractory depression is diagnosed according to the diagnosis result output by the classification model.

[0032] The present invention uses important inflammatory factors of TRD to screen characteristic genes that are ultimately used to input the classification model, thereby effectively utilizing blood gene expression data to achieve low-cost, high-efficiency, and high-accuracy TRD diagnosis, providing convenience for the diagnosis of a large number of patients and having broad application prospects. In order to better illustrate the purpose, technical solutions, and advantages of the present invention, the present invention will be further described below in conjunction with specific embodiments.

[0033] The present invention provides a method for diagnosing refractory depression using a combination of blood gene expression, by using a trained classification model to diagnose TRD patients according to the output of the classification model; the trained machine classification model is based on a training data set, with a blood differential gene expression matrix as input and a predicted diagnosis result as output, and is trained by minimizing the error between the output of the classification model and the TRD label as the goal. In general, the classification model is first trained, such as Figure 1 As shown, the training method of the classification model is as follows:

[0034] (1) Constructing a training data set, wherein each sample of the training data set includes a blood differential gene expression matrix and a diagnostic label of a TRD patient or a non-patient, wherein the TRD label is set according to the diagnosis result, wherein the patient diagnosed with TRD is set as the TRD label, and the normal person is set as the non-TRD label.

[0035] The blood differential gene expression matrix is ​​based on the differential genes obtained by screening the blood transcriptome of TRD patients and normal people, and further based on the machine model calculation weight screening to obtain differential genes with weights greater than the weights of important inflammatory factors of TRD; important inflammatory factors of TRD are one or more of IL-6, IL-1β and TNF-α; in a specific implementation case, TRD-related literature and blood transcriptome data of TRD and normal people can be searched from the GEO (https: / / www.ncbi.nlm.nih.gov / geo / ) database, and the TRD label can be set according to the diagnosis results. Then, TRD is used as the experimental group and normal people are used as the control group for differential gene analysis, and the first group of genes is obtained as differential genes by screening (for example, the threshold is P-value≤0.05, |log2FC|≥log2(1.2)). In a preferred embodiment, KEGG and GO-BP enrichment analysis can also be performed on the obtained up- and down-regulated differential genes, that is, the first group of differential genes, respectively, and the number of genes is reduced by enrichment analysis, and the second group of genes is obtained by screening as differential genes. In a preferred embodiment, since the data comes from different platforms, the collected data can also be used to remove batch effects on the data from different platforms using the limma package removebatcheffect in the R language.

[0036] Furthermore, based on the weights calculated by the machine model, the differentially expressed genes with weights greater than the weights of the important inflammatory factors of TRD are screened; the important inflammatory factors of TRD are one or more of IL-6, IL-1β and TNF-α. In a preferred embodiment, the first group of genes or the second group of genes screened can be trained using a random forest or a support vector machine, and the model feature coefficient (coef_) is used in the svm to obtain the weight corresponding to each gene feature, and the random forest uses the feature importance (feature_importances_) to obtain the weight.

[0037] (2) constructing a machine classification model, taking the blood differential gene expression matrix of the training data set as input, and training by minimizing the error between the output of the classification model and the TRD label as the goal to obtain a trained classification model. The classification model of the present invention can adopt a conventional machine learning model, such as a classification model based on a support vector machine, a classification model based on a random forest, and / or a one-dimensional convolutional neural network.

[0038] Finally, the trained machine classification model can be used to diagnose refractory depression using a combination of blood gene expression, namely:

[0039] Obtaining a blood differential gene expression matrix of the patient to be tested;

[0040] The blood differential gene expression matrix is ​​input into a trained machine classification model, and whether the patient to be tested suffers from refractory depression is diagnosed according to the diagnosis result output by the classification model.

[0041] The effect of the present invention is further described below in conjunction with a specific embodiment:

[0042] A method for diagnosing refractory depression using a blood gene expression panel, specifically comprising the following steps:

[0043] (1) Constructing a training data set, which includes the following sub-steps:

[0044] (1.1) TRD-related blood transcriptome literature and data collection:

[0045] The embodiment of the present invention searches TRD-related literature and blood transcriptome data of TRD and normal people from the GEO (https: / / www.ncbi.nlm.nih.gov / geo / ) database, and sets the TRD label according to the diagnosis results. The collected data are shown in Table 1, where TRD is the blood transcriptome data corresponding to patients with refractory depression, and CONTROL is the blood transcriptome data corresponding to normal people.

[0046] Table 1: Statistics of blood transcriptome data

[0047] Gene expression profiles predict response to infliximab in patients with treatment-resistant depression Rapid antidepressant response to ketamine in patients with treatment-resistant depression is associated with a whole-blood transcriptional signature Blood gene expression in major depressive disorder Gene expression profiles of comorbid PTSD and depression Gene expression analysis of whole blood samples obtained from donors diagnosed with major depressive disorder and healthy controls

[0048] Label GSE45468 GSE185855 GSE19738 GSE67663 GSE98793 total TRD(1) 58 47 0 0 0 105 CONTROL(0) 0 22 34 70 64 190

[0049] (1.2) Perform differential analysis on transcriptome data after removing batches:

[0050] Furthermore, since the data come from different platforms, it is necessary to use the R language limma package removebatcheffect to remove the batch effect of the collected data on the data of different platforms. Differential gene analysis was performed with TRD as the experimental group and normal people as the control group. The screening threshold was P-value≤0.05, |log2FC|≥log2(1.2). A total of 582 differential genes were screened, and KEGG and GO-BP enrichment analysis were performed on the obtained up- and down-regulated differential genes, respectively. A total of 122 KEGG pathways and 702 GO pathways were screened. Table 2 shows the top 5 up-regulated differential genes and the top 5 down-regulated differential genes, the top 5 KEGG enrichment analysis pathways and the top 5 GO-BP pathway analysis results.

[0051] Table 2: Five up-regulated and down-regulated genes and pathways

[0052]

[0053]

[0054] (1.3) RF and SVM calculate differential gene weights:

[0055] RF and SVM are used to calculate the gene weight value, and random forest and support vector machine are used to train all features. In svm, the model feature coefficient (coef_) is used to obtain the weight corresponding to each feature, and random forest uses feature importance (feature_importances_) to obtain the weight. Figure 2 and Figure 3 The RF and SVM models calculate the different proportions of each gene in the model training.

[0056] (1.4) Among them, the three important inflammatory factors related to TRD, IL-6, IL-1β and TNF-α (in TRD patients, the expression of these three inflammatory factors is upregulated, leading to the further release of proinflammatory cytokines, inducing inflammatory depression) will be used as weight screening indicators, and the weighted genes with weights before all or one of the three weights will be selected for training).

[0057] In addition, this embodiment selected 47 samples as the test set, of which 21 were CONTROL and 26 were TRD patients.

[0058] (2) Construct a machine classification model, take the gene expression matrix of the training data set as input, and train by minimizing the error between the output of the classification model and the TRD label to obtain a trained classification model. The classification model of the present invention can adopt a conventional machine learning model, and in this embodiment, it includes:

[0059] Classification model built and trained based on support vector machine: Support vector machine is a supervised learning method that can be widely used in statistical classification and regression analysis. Support vector machine maps vectors to a higher dimensional space and establishes a maximum margin hyperplane in this space. Two parallel hyperplanes are built on both sides of the hyperplane that separates the data. The separating hyperplane maximizes the distance between the two parallel hyperplanes. It is assumed that the greater the distance or gap between parallel hyperplanes, the smaller the total error of the classifier. Select the kernel function "Linear, Sigmod and Poly", and then use the Python function GridSearchCV to determine the penalty parameter C. The parameter is selected from 0.00001 to 1000, with a multiple of 10. Use the support vector machine model in the Python-based machine learning module library Scikit-learn to train and predict the samples. Select the best model and parameter C based on AUC. Train according to the training data set.

[0060] Classification model based on random forest: Random forest is a classifier that uses multiple decision trees to train and predict samples. The number of decision trees greatly affects the accuracy of the algorithm, so the Python function GridSearchCV is used to determine the number of decision trees parameter n_estimators, with a parameter selection range of 0 to 500 and a step size of 10. The random forest classifier in the Python-based machine learning module library Scikit-learn is used to train and predict samples. The best model and parameter estimators are selected based on the highest AUC. Training is performed based on the training data set.

[0061] A one-dimensional convolutional neural network is a calculation applied to two number arrays (one input, one kernel). The process can be simply described as follows: first, align the two rows of numbers to the left, multiply the corresponding numbers above and below, and then sum them to get a new value; then slide the kernel to the right according to the stride, multiply and sum them again to get a new value; continue the sliding and calculation process until the right end of the kernel exceeds the input array, at which point no more calculation is done and the training data set is input for training.

[0062] In this example, in order to illustrate the excellent effect of the present invention, different differentially expressed genes were selected according to log2FC for parameter optimization and model training as a control.

[0063] Finally, all models built with the best parameters using the test set are evaluated using four scoring indicators: accuracy, precision, recall, and F1 score. Their calculation formulas are as follows, where TP is true positive, TN is true negative, FP is false positive, and FN is false negative.

[0064]

[0065] According to logFC or weight, different differentially expressed genes training sets were selected and input into the model. Among them, according to log2FC, the top 20 up- and down-regulated 40 differentially expressed genes were selected as training sets, and then the GridSearchCV function was used to select the number of RF trees parameter n_estimators, and the best AUC score was 0.86. For support vector machines (SVM), we used different training sets and the GridSearchCV function to select the optimal parameters of the SVM algorithm penalty coefficient parameter C and the gamma model. We found that the best performance model can be obtained by using the features with a weight value greater than TNF-α as training. The best AUC scores for different kernel functions were 0.95 for Linear, 0.93 for Sigmod, and 0.95 for Poly, and the model diagnosis results were significantly improved. For the one-dimensional convolutional model (1D CNN), we optimized the model by changing the learning rate (lr). The best training set was also based on the features with a weight value greater than TNF-α as training, and the best AUC score was 0.92. Figure 4 is the AUC score obtained with the best training set and best parameters.

[0066] Figure 5 The results are obtained by selecting the top 40 differentially regulated genes ranked in the top 20 in terms of up- and down-regulation among the differentially regulated genes according to log2FC as the training set based on RF, and selecting the features with weight values ​​greater than those before TNF-α as the optimal parameters and optimal gene set for training based on SVM and one-dimensional convolutional neural network models according to the present invention; Figure 5 As shown, the TRD diagnosis classification models constructed by the present invention have good classification performance. The trained classification models can be used to diagnose existing TRD patients.

[0067] An electronic device provided by an embodiment of the present invention includes one or more processors, which are used to implement a method for diagnosing refractory depression using a blood gene expression combination in the above embodiment.

[0068] The electronic device of the present invention can be applied to any device with data processing capability, and the device with data processing capability can be a device or apparatus such as a computer.

[0069] The device embodiment can be implemented by software, hardware or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, it is formed by the processor of any device with data processing capability in which it is located to read the corresponding computer program instructions in the non-volatile memory into the memory for execution. From the hardware level, in addition to the processor, memory, network interface, and non-volatile memory, any device with data processing capability in which the device in the embodiment is located can also include other hardware according to the actual function of the device with data processing capability, which will not be described in detail.

[0070] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0071] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of the present invention. Ordinary technicians in this field can understand and implement it without paying creative work.

[0072] An embodiment of the present invention further provides a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, a method for diagnosing refractory depression using a blood gene expression combination in the above embodiment is implemented.

[0073] The computer-readable storage medium may be an internal storage unit of any device with data processing capability described in any of the aforementioned embodiments, such as a hard disk or a memory. The computer-readable storage medium may also be any device with data processing capability, such as a plug-in hard disk, a smart media card (SMC), an SD card, a flash card, etc. equipped on the device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of any device with data processing capability and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capability, and may also be used to temporarily store data that has been output or is to be output.

[0074] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.

Claims

1. A method for diagnosing refractory depression using a blood gene expression panel, characterized in that: include: Obtaining a blood differential gene expression matrix of the patient to be tested; The blood differential genes are differential genes obtained by screening the blood transcriptomes of TRD patients and normal people, and further screened based on the weights calculated by the machine model, and the weights of the differential genes are greater than the weights of important inflammatory factors of TRD; the important inflammatory factors of TRD are one or more of IL-6, IL-1β and TNF-α; the blood differential gene expression matrix is ​​input into a trained machine classification model, and the patient to be tested is diagnosed whether he suffers from refractory depression according to the diagnosis results output by the classification model; The trained machine classification model is based on a training data set, takes a blood differential gene expression matrix as input, takes a predicted diagnosis result as output, and is trained by minimizing the error between the output of the classification model and the TRD label as the goal.

2. The method according to claim 1, characterized in that The differentially expressed genes obtained by screening the blood transcriptomes of TRD patients and normal subjects were obtained by screening using the following method: Collect blood transcriptome data from TRD and normal subjects; Differential gene analysis was performed with TRD as the experimental group and normal subjects as the control group, and the first group of genes obtained by screening were the differential genes.

3. The method according to claim 2, characterized in that The screening threshold was P-value≤0.05,|log2FC|≥log2(1.2).

4. The method according to claim 2, characterized in that: It also includes further KEGG and GO-BP enrichment analysis after the first group of genes are screened, and the second group of genes obtained by screening are differential genes.

5. The method according to any one of claims 2 to 4, characterized in that: The machine model is a random forest or a support vector machine; wherein, in the support vector machine, the blood differential gene expression matrix composed of the first group of genes or the second group of genes obtained by screening is used as input for training, and the weight corresponding to each gene feature is obtained using the model feature coefficient; in the random forest, the blood differential gene expression matrix composed of the first group of genes or the second group of genes obtained by screening is used as input for training, and the weight corresponding to each gene feature is obtained using the feature importance.

6. The method according to claim 1, characterized in that Each sample of the training data set includes a blood differential gene expression matrix and a corresponding TRD label.

7. The method according to claim 1, characterized in that The machine classification model includes a classification model constructed based on a support vector machine, a classification model constructed based on a random forest and / or a one-dimensional convolutional neural network.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for diagnosing refractory depression using a blood gene expression combination as described in any one of claims 1 to 7 is implemented.

9. A storage medium comprising computer executable instructions, which, when executed by a computer processor, implement a method for diagnosing refractory depression using a blood gene expression panel as described in any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of a method for diagnosing refractory depression using a blood gene expression combination as described in any one of claims 1 to 7 are implemented.

Images