Training method of respiratory tract infection disease progress and prognosis prediction model

Abstract

The invention relates to a training method of a respiratory tract infection disease progress and prognosis prediction model. The training method comprises the following steps: extracting mRNA from peripheral blood of a target patient, and carrying out transcriptome sequencing to obtain a sequencing result; based on the ferroptosis related gene set, comparing ferroptosis score differences of two groups of patients with community-acquired pneumonia and sepsis, and screening corresponding ferroptosis related genes with statistical significance from a sequencing result; screening out genes meeting preset conditions from the ferroptosis related genes based on LASSO regression; and establishing an RTI clinical outcome prediction model through logistic regression by taking whether the patient is sepsis or not as an outcome dichotomy variable and taking the screened gene expression quantity as a prediction variable. According to the invention, after the prediction model is subjected to machine learning screening such as LASSO and the like, the core feature with the highest prediction value is reserved, so that the risk of over-fitting of the model on training data is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine learning, in particular to a training method of a respiratory tract infection disease progression and prognosis prediction model. BACKGROUND

[0002] Community-acquired pneumonia (CAP) remains an important cause of human death worldwide, causing a significant burden to the economy and society. When it develops into acute respiratory distress syndrome (ARDS) or sepsis, the mortality rate increases greatly. Although the pathophysiological mechanism of ARDS has revealed its close relationship with alveolar epithelial injury, abnormal vascular permeability and inflammation, there is still no reliable early biomarker that can accurately predict the progression and outcome of respiratory tract infection.

[0003] Transcriptomics systematically analyzes the gene expression profile of cells under specific physiological or pathological conditions through RNA sequencing technology, revealing the dynamic changes of gene regulatory networks and functional pathways. Foreign countries have developed gene markers to identify respiratory viral and bacterial infections based on host transcriptomic information, and established gene markers to predict the prognosis of sepsis patients based on public databases. The above markers all establish identification or prognosis prediction models through peripheral blood gene expression analysis of the host, however, the above gene markers still have the following defects and deficiencies:

[0004] (1) The data mainly comes from public databases, and the study cannot control the patient's demographic characteristics, infection type, disease severity and individual treatment differences. These heterogeneities may affect the performance and generalization ability of the model.

[0005] (2) The 30-day mortality rate is used as the endpoint indicator in the study of sepsis mortality prediction models, which is a major measure of clinical outcomes and may overlook other important intermediate endpoints, such as the progression of community-acquired pneumonia patients to sepsis or ARDS, and related models still need to be developed and verified.

[0006] In addition, the current clinical assessment of sepsis patient disease severity is through SOFA score, and the assessment of pneumonia patient disease severity is through CURB-65, but in clinical practice, it has the following limitations:

[0007] (1) SOFA score and the like need various test data for sequential evaluation, which is more suitable for inpatients, but has certain limitations for early diagnosis and rapid diagnosis of emergency patients; since it is a rule-based technology that simplifies multidimensional data into a single scalar value, it may lead to inaccurate classification or clustering of patient status.

[0008] (2)CURB-65 divides patients into two categories of severe and non-severe, and cannot accurately identify patients with low mortality risk who are suitable for early discharge or home treatment. In previous studies, this score underestimated the severity of young patients and did not include factors such as hypoxemia that require oxygen therapy. SUMMARY

[0009] To solve the above problems, the purpose of the embodiment of the present application is to provide a training method of a respiratory tract infection disease progression and prognosis prediction model.

[0010] A training method of a respiratory tract infection disease progression and prognosis prediction model, comprising:

[0011] Step 1: Collect target patients into community-acquired pneumonia group and sepsis group;

[0012] Step 2: Extract mRNA from the peripheral blood of the target patients for transcriptome sequencing to obtain sequencing results;

[0013] Step 3: Based on the iron death related gene set, compare the iron death scores of the two groups of patients with community-acquired pneumonia and sepsis, and screen out corresponding iron death related genes with statistical significance from the sequencing results;

[0014] Step 4: Based on LASSO regression, screen out genes that meet the preset conditions from the iron death related genes;

[0015] Step 5: Take whether the patient is sepsis as the outcome binary variable, and the expression amount of the screened gene as the prediction variable, and establish an RTI clinical outcome prediction model by logistic regression;

[0016] Step 6: Use the RTI clinical outcome prediction model to complete the prediction of the respiratory tract infection patients to obtain the prediction results.

[0017] Preferably, step 3: based on the iron death related gene set, comparing the iron death scores of the two groups of patients with community-acquired pneumonia and sepsis, and screening out corresponding iron death related genes with statistical significance from the sequencing results, comprises:

[0018] Step 3.1: Standardizing the count data of the iron death related genes to obtain a standardization factor;

[0019] Step 3.2: Using the standardization factor to perform negative binomial regression on the count data of the iron death related genes to obtain a negative binomial distribution model;

[0020] Step 3.3: Based on the negative binomial distribution model, the Wald test is used to compare the difference in gene expression in the iron death related gene set and the sequencing results, and genes with P values within the set range are selected as statistically significant corresponding iron death related genes.

[0021] Preferably, in step 3.1, the formula is used:

[0022]

[0023] The count data of the iron death related genes are standardized to obtain a standardization factor; wherein s i represents the standardization factor, median j represents the median of all genes, k ji is the count data of gene j in sample i.

[0024] Preferably, step 4: Based on LASSO regression, genes that meet the preset conditions are selected from the iron death related genes, including:

[0025] The LASSO regression is performed on the expression of the iron death related genes as the characteristic vector and whether the patient is sepsis as the outcome variable to obtain genes that meet the preset conditions; wherein the loss function of the LASSO regression is:

[0026]

[0027] wherein y i is the clinical outcome of the i-th patient, which is a binary variable indicating whether the patient is sepsis, y i = 1 indicates sepsis, and y i = 0 indicates non-sepsis; is the predicted value of the i-th patient, β is the regression coefficient vector, n is the number of patients, λ is the regularization parameter, and ||β||1 is the L1 norm of β.

[0028] Preferably, step 5: Taking whether the patient is sepsis as the outcome binary variable and the expression of the screened genes as the prediction variable, an RTI clinical outcome prediction model is established by logistic regression, including:

[0029] Taking whether the patient is sepsis as the outcome binary variable and the expression of the screened genes as the prediction variable, an RTI clinical outcome prediction model is established by logistic regression; wherein the logistic regression algorithm is:

[0030] logit(P(Y = 1)) = β0 + β1X1

[0031] wherein, is a logit transformation that maps the probability P to the real number domain, β0 is an intercept term, β1 is a first regression coefficient, Y is a sepsis outcome variable, and X1 is a PGD expression in peripheral blood of an acute respiratory infection patient.

[0032] Preferably, in step 5, further comprising:

[0033] Step 5.1: evaluating the RTI clinical outcome prediction model to obtain a plurality of evaluation indexes;

[0034] Step 5.2: normalizing each evaluation index to obtain a normalized evaluation index;

[0035] Step 5.3: determining a weighting coefficient according to a proportion of each evaluation index in all evaluation indexes;

[0036] Step 5.4: weighted summing the normalized evaluation indexes to obtain a comprehensive evaluation index;

[0037] Step 5.5: when the comprehensive evaluation index is within a preset range, the RTI clinical outcome prediction model training is completed.

[0038] Preferably, in step 5.3, the formula is adopted:

[0039]

[0040] determining a weighting coefficient according to a proportion of each evaluation index in all evaluation indexes; wherein, P i represents the proportion of the i-th evaluation index, x i represents the normalized value of the i-th evaluation index, m represents the number of evaluation indexes, and w i represents the weight of the i-th index.

[0041] The application further provides an electronic device, including a bus, a transceiver, a memory, a processor and a computer program stored in the memory and executable on the processor, the transceiver, the memory and the processor being connected through the bus, characterized in that the computer program is executed by the processor to realize the steps in the training method of the respiratory infection disease progression and prognosis prediction model.

[0042] The application further provides a computer readable storage medium, which stores a computer program, characterized in that the computer program is executed by a processor to realize the steps in the training method of the respiratory infection disease progression and prognosis prediction model.

[0043] According to the specific embodiments of the application, the following technical effects are achieved:

[0044] The present application relates to a kind of respiratory infection disease progression and prognosis prediction model training method, compared with prior art, the present application is still lack of the problem that can accurately predict RTI patient disease progression and prognosis prediction model, by including different severity of respiratory infection patient, by transcriptome sequencing and machine learning, the gene related to disease severity is screened and non-zero coefficient gene is obtained, then a model for predicting RTI clinical outcome by single gene is established by logistic regression, after the model is screened by LASSO etc. Machine learning, the core feature with most predictive value is retained, the risk of overfitting is reduced, and the prediction performance of the model is improved.

[0045] In order to make the above-mentioned objects, features and advantages of the present application more apparent and easy to understand, the following preferred embodiments are described in detail below, and the accompanying drawings are described as follows. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0047] Figure 1 A respiratory infection disease progression and prognosis prediction model training method flow chart is provided for the present application;

[0048] Figure 2 RTI patient iron death score is provided for the present application;

[0049] Figure 3 The coefficient path diagram of LASSO regression of RTI clinical outcome prediction model provided by the present application;

[0050] Figure 4 The LASSO regression coefficient of RTI clinical outcome prediction model provided by the present application;

[0051] Figure 5 The performance of RTI clinical outcome prediction model provided by the present application in test set;

[0052] Figure 6 The ROC curve of RTI clinical outcome prediction model in validation group queue 1 provided by the present application;

[0053] Figure 7 The ROC curve of RTI clinical outcome prediction model in validation group queue 2 provided by the present application;

[0054] Figure 8 The decision curve of RTI clinical outcome prediction model provided by the present application. DETAILED DESCRIPTION

[0055] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, which is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0056] In addition, the terms "first", "second" are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise explicitly specified and limited.

[0057] In the present application, unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection", "fixing" and the like should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0058] Please refer to Figure 1 A training method of a respiratory tract infection disease progression and prognosis prediction model, comprising:

[0059] Step 1: Collect target patients into community-acquired pneumonia group and sepsis group;

[0060] Step 2: Extract mRNA from the peripheral blood of the target patient to perform transcriptome sequencing to obtain sequencing results;

[0061] In practical application, the present application collects patients into learning group according to the following standards in the acute respiratory infection cohort of the Department of Respiratory and Critical Care Medicine of Peking Union Medical College Hospital, and divides them into community-acquired pneumonia group and sepsis group.

[0062] Community-acquired pneumonia group:

[0063] ① Age ≥ 18 years old;

[0064] ② Sign the informed consent form;

[0065] ③According to the researchers' judgment, can follow the study visit;

[0066] ④Meet the diagnostic criteria of community-acquired pneumonia in Chinese adult community-acquired pneumonia diagnosis and treatment guidelines (2016 edition);

[0067] ⑤The onset is within 8 days of enrollment.

[0068] Sepsis group:

[0069] ①Age ≥18 years;

[0070] ②Signed informed consent;

[0071] ③According to the researchers' judgment, can follow the study visit;

[0072] ④Meet the diagnostic criteria of sepsis 3.0.

[0073] Exclusion criteria:

[0074] ①There is respiratory failure related to cardiopulmonary underlying disease;

[0075] ②Severe liver and kidney dysfunction, pregnancy or lactation;

[0076] ③Meet the indications for organ transplantation or have undergone organ transplantation surgery;

[0077] ④Known HIV infection;

[0078] ⑤Unstable angina pectoris or myocardial infarction within 30 days before screening and without vascular recanalization treatment, massive cerebral hemorrhage;

[0079] ⑥The researcher judged that it was not suitable for enrollment.

[0080] The process of peripheral blood transcriptome sequencing of patients is as follows:

[0081] 1. Collect 2 mL of peripheral blood, and separate peripheral blood mononuclear cells (PBMCs) within 4 hours using Ficoll-Paque density gradient centrifugation, add TRIzol, freeze at -80℃, and store in liquid nitrogen for standby.

[0082] 2. Extract RNA and systematically evaluate RNA integrity, purify mRNA from total RNA to construct a library, and perform transcriptome sequencing by Illumina NovaSeq 6000.

[0083] 3. Align the sequencing results with the reference genome GRCh38 by Hisat2, and use Featurecounts for gene counting.

[0084] Step 3: comparing the difference of ferroptosis score between community-acquired pneumonia and sepsis patients based on the set of ferroptosis-related genes, and screening the corresponding ferroptosis-related genes with statistical significance from the sequencing results;

[0085] In step 3, it includes:

[0086] Step 3.1: standardizing the count data of ferroptosis-related genes to obtain a standardization factor;

[0087] Step 3.2: using the standardization factor to perform negative binomial regression on the count data of ferroptosis-related genes to obtain a negative binomial distribution model;

[0088] Step 3.3: comparing the difference of gene expression between the set of ferroptosis-related genes and the sequencing results based on the negative binomial distribution model using Wald test, and screening the genes with P value within the set range as the corresponding ferroptosis-related genes with statistical significance.

[0089] In step 3, first, the set of ferroptosis-related genes is collected from previously published literature, and the difference of ferroptosis score between community-acquired pneumonia and sepsis patients is compared (P<0.001).

[0090] Using the screened differential genes, the process is as follows:

[0091] ① Based on the negative binomial distribution model, the count data of genes are standardized to obtain a size factor to eliminate the influence of different sample sequencing depth differences;

[0092] The standardization process is as follows:

[0093] A: standardizing between samples (i.e., the gene expression matrix of the transcriptome sequencing of acute respiratory infection patients), calculating the standardization factor s of each sample to adjust the sequencing depth difference of different samples. Its formula is: i , where median j represents the median of gene dimension (i.e., all genes), and k ji represents the count data of gene j in sample i.

[0094] B: modeling the count of each gene j in sample i as a negative binomial distribution: K ji ~ NB(μ ji , α j ), where the mean function μ ji = q j s i , and q j is the relative expression of the gene.

[0095] ② Estimate the relationship between the mean and variance of gene expression.​

[0096] The variance function of the negative binomial distribution model is adopted, i.e., the variance is equal to the mean plus the square of the mean multiplied by the dispersion coefficient: Var(K ji ) = μ ji + μ ji 2 α j ;

[0097] ③ The gene expression and the dispersion coefficient are estimated by maximum likelihood estimation and the like.

[0098] For each gene j, a log-likelihood function is constructed: where k ij is the count of gene j in sample i, and P(K ij = k ij | μ ji , α j ) is the probability mass function of the negative binomial distribution. By maximizing this log-likelihood function, i.e., taking the partial derivatives of μ j and α j and setting them equal to zero, and solving the equation set to estimate the parameters μ j and α j . For the estimation of μ j , it is usually necessary to combine the size factor s i of the sample, the mean μ ji = q j s i , q j where is the relative expression of the gene.

[0099] ④ The Wald test is used to compare the differences in gene expression between different groups, and the P value corrected by the Benjamin-Hochberg method is defined as P < 0.05 to have statistical significance. The clusterProfiler package is used for GO and KEGG enrichment analysis to compare the up-regulation of ferroptosis between the two groups, and the ferroptosis score of each group is calculated, as shown in Figure 2 .

[0100] Step 4: Screening genes from ferroptosis-related genes that meet the preset conditions based on LASSO regression;

[0101] In step 4, it includes:

[0102] ① The edge package in R language is used to standardize the gene count data to the CPM value of the gene by logCPM, and the normalized data of the expression matrix is obtained;

[0103] * CPM calculation formula: CPM = A / mapped reads x 10 6

[0104] where A is the read count mapped to a certain gene, and mapped reads is the total reads mapped to the reference transcriptome in the sample.

[0105] ② Set the random seed number to divide the learning set into training set (70%) and test set (30%) with sepsis as the outcome variable and the expression of iron death related genes as the feature vector.

[0106] ③ Use the glmnet package in R language to input the feature vector and outcome variable of the training set data, calculate a set of λ values based on cross-validation, and fit the corresponding model. With the change of the regularization parameter λ, the regression coefficients of some genes are gradually compressed to zero, realizing gene screening.

[0107] Regularization parameter: λ is a non-negative number that controls the strength of regularization, used to balance the relationship between model fitting and complexity. When λ increases, the influence of the regularization term also increases, which tends to compress more regression coefficients to zero, making the model simpler (i.e. variable selection); when λ decreases, the model is closer to ordinary least squares regression, allowing more variables to enter the model. The loss function formula of LASSO regression:

[0108]

[0109] where, is the mean of the residual sum of squares, used to measure the fitting degree of the model to the data. yi is the true value of the i-th sample (i.e. the clinical outcome of the i-th patient, which is a binary variable indicating whether the patient is sepsis, for example, 1 represents sepsis, and 0 represents non-sepsis), is the predicted value of the i-th sample, where x i is the independent variable vector containing the expression levels of multiple genes, β is the regression coefficient vector, used to measure the influence of each gene on the clinical outcome, n is the sample size (i.e. the sample size of the training set of the learning set), λ||β||1 is the L1 regularization term, used to control the complexity of the model to prevent overfitting, ||β||1 is the L1 norm of β, i.e. where p is the number of features (i.e. the number of genes in the iron death gene set).

[0110] Step 5: Establish the RTI clinical outcome prediction model by logistic regression with sepsis as the outcome binary variable and the expression of the screened genes as the predictor variable.

[0111] In step 5, as Figures 3-4As shown, based on the key genes screened by LASSO regression, a prediction model is further constructed. The gene expression data screened is used as the independent variable, and whether the patient is sepsis is used as the target variable. Based on different regularization parameters λ, the corresponding regression coefficients are extracted, and the genes corresponding to the non-zero coefficients are selected genes used to construct the prediction model. Finally, one gene (PGD) is selected as the prediction model, and the process is as follows:

[0112] ①Using the single gene variable (PGD) obtained by LASSO regression, in the training set, whether the patient is sepsis is used as the outcome binary variable (dependent variable), and the PGD expression is used as the prediction variable (independent variable). A clinical prediction tool is established by logistic regression;

[0113] Logistic regression algorithm:

[0114] logit(P(Y=1))=β0+β1X1+β2X2+…+β p X p

[0115] Wherein, is the logit transformation, which maps the probability P to the real number domain. The output of this linear combination is mapped to the interval (0, 1), which represents the probability. The threshold is set to 0.5, and the probability is converted to a category label: if P≥0.5, predict y=1 (i.e. predict sepsis); if P<0.5, predict y=0 (i.e. predict non-sepsis), β0 is the intercept term, β1, β2, β p is the regression coefficient, which represents the contribution of each independent variable to the dependent variable, X p is the expression of a specific gene.

[0116] The algorithm in the present application is:

[0117] logit(P(Y=1))=β0+β1X1

[0118] Wherein, Y is the sepsis outcome variable, and X1 is the PGD expression in the peripheral blood of patients with acute respiratory tract infection.

[0119] ②Model validation is performed in the test set, and the prediction accuracy, sensitivity, specificity and other indicators of the model are calculated, and the receiver operating characteristic (ROC) curve is drawn, and the area under the ROC curve (AUC) is calculated. AUC 0.6-0.8 is considered to have moderate discrimination ability, and AUC>0.8 is considered to be good.

[0120] As Figure 5 shown, after analysis, the AUC of the RTI clinical outcome prediction model in the test set of the learning group is 0.92, and the prediction ability is good. At this time, the sensitivity and specificity are 0.78 and 0.90 respectively, and the positive predictive value and negative predictive value are 0.64 and 0.95 respectively.

[0121] A patient group was collected in the acute respiratory infection cohort in China-Japan Friendship Hospital according to the same inclusion and exclusion criteria as the learning group, and the patients were divided into community-acquired pneumonia patients and sepsis patients according to the aforementioned diagnostic criteria. The gene expression matrix of the RTI clinical outcome prediction model was obtained according to the transcriptome sequencing results of the patients. The ROC curve was drawn according to the prediction model, and the AUC, sensitivity, specificity, and positive and negative predictive values were calculated.

[0122] As shown in Figure 6 , after analysis, the AUC of the RTI clinical outcome prediction model in the validation group cohort 1 was 0.82, and the prediction ability was good. At this time, the sensitivity and specificity were 0.88 and 0.51, respectively, and the positive and negative predictive values were 0.65 and 0.80, respectively.

[0123] The acute respiratory infection cohort in the emergency department of Peking Union Medical College Hospital was used as the validation cohort 2, and the severe pneumonia patients were divided into survival group and death group. According to the peripheral blood transcriptome sequencing results, the expression matrix of the prediction model gene was obtained, the ROC curve was drawn, and the AUC was calculated to evaluate the ability of the model to predict whether the severe pneumonia patient died, and the predicted sensitivity and specificity and positive and negative predictive values were calculated.

[0124] As shown in Figure 7 , after analysis, the AUC of the RTI clinical outcome prediction model in the validation group cohort 2 was 0.86, and the prediction ability was good. At this time, the sensitivity and specificity were 0.47 and 1, respectively, and the positive and negative predictive values were 1 and 0.59, respectively.

[0125] In addition to the above model evaluation content, decision curves were drawn to guide clinical decision-making to comprehensively evaluate the overall performance of the model. As shown in Figure 8 - A and Figure 8 B are the decision curves in the test set of the learning group and the validation group cohort 1, respectively. The horizontal coordinate represents the risk threshold, and the vertical coordinate represents the net benefit, wherein the gray line (All) represents the standardized net benefit of all patients receiving the RTI clinical outcome prediction model to predict disease progression or clinical outcome, and the black horizontal line (None) represents all patients not receiving the prediction of the RTI clinical outcome prediction model. After analysis, it was found that in the test set, the model curve was above the All or None curve in almost all threshold ranges, indicating that disease progression prediction according to the RTI clinical outcome prediction model can bring clinical net benefit; in the validation group cohort 1, the model can bring benefit to predict the clinical outcome of the patients when the threshold probability is higher than 0.6, and the part below 0.6 almost overlaps with the gray line. In clinical application, appropriate threshold probability should be selected to apply the prediction model in combination with specific clinical background, patient risk preference and other factors.

[0126] In step 5, the application can also weight the overall performance of the index comprehensive evaluation model, including:

[0127] Step 5.1: The RTI clinical outcome prediction model is evaluated to obtain a plurality of evaluation indexes; such as prediction accuracy, sensitivity, specificity and the like.

[0128] Step 5.2: Each evaluation index is normalized to obtain a normalized evaluation index;

[0129] Step 5.3: Determine the weighting coefficient according to the proportion of each evaluation index in all evaluation indexes;

[0130] In step 5.3, the formula is:

[0131]

[0132] Determine the weighting coefficient according to the proportion of each evaluation index in all evaluation indexes; wherein, P i represents the proportion of the i-th evaluation index, x i represents the normalized value of the i-th evaluation index, m represents the number of evaluation indexes, w i represents the weight of the i-th index.

[0133] Step 5.4: Weighted sum of the normalized evaluation index to obtain a comprehensive evaluation index;

[0134] Step 5.5: When the comprehensive evaluation index is in the preset range, the RTI clinical outcome prediction model training is completed.

[0135] After weighting a plurality of indexes, the application can complement and correct each other, thereby reducing the influence of single index deviation and improving the robustness of the model.

[0136] The application also provides an electronic device, including a bus, a transceiver, a memory, a processor and a computer program stored on the memory and executable on the processor, the transceiver, the memory and the processor are connected through the bus, characterized in that the computer program is executed by the processor to realize the steps of the above-mentioned respiratory tract infection disease progress and prognosis prediction model training method, and the beneficial effects of the above-mentioned respiratory tract infection disease progress and prognosis prediction model training method are the same, and here is not repeated.

[0137] The application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the training method of the respiratory tract infection disease progression and prognosis prediction model.

[0138] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or alternative technical solutions within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A training method for a prediction model of the progression and prognosis of respiratory infectious diseases, characterized in that, The method comprises the following steps: Step 1: Collect target patients and divide them into a community-acquired pneumonia group and a sepsis group; Step 2: Extract mRNA from peripheral blood of the target patients to obtain sequencing results through transcriptome sequencing; Step 3: Compare the iron death scores of the patients in the community-acquired pneumonia group and the sepsis group based on an iron death-related gene set, and screen corresponding iron death-related genes with statistical significance from the sequencing results; Step 4: Screen genes meeting preset conditions from the iron death-related genes based on LASSO regression; Step 5: Take whether a patient is sepsis as a binary outcome variable, and take the expression of the screened genes as a prediction variable, and establish an RTI clinical outcome prediction model through logistic regression; Step 6: Use the RTI clinical outcome prediction model to complete prediction on patients with respiratory tract infections to obtain prediction results. 2.The training method of a respiratory infection disease progression and prognosis prediction model according to claim 1, characterized in that, The step 3 comprises the following steps: Step 3.1: Standardize the count data of the iron death-related genes to obtain a standardization factor; Step 3.2: Perform negative binomial regression on the count data of the iron death-related genes by using the standardization factor to obtain a negative binomial distribution model; Step 3.3: Compare the differences between the iron death-related gene set and the gene expression in the sequencing results based on the negative binomial distribution model by using Wald test, and screen genes with P values in a set range as corresponding iron death-related genes with statistical significance. 3.The training method of a respiratory infection disease progression and prognosis prediction model according to claim 2, characterized in that, In step 3.1, the formula is as follows: The count data of the ferroptosis-related genes are standardized to obtain a standardization factor; wherein, s i represents the standardization factor, median j represents that the median is calculated for all genes, k ji count data of the gene j in the sample i. 4.The training method of a respiratory infection disease progression and prognosis prediction model according to claim 3, characterized in that, Step 4: Screen genes meeting preset conditions from the iron death-related genes based on LASSO regression, which comprises the following steps: Take whether a patient is sepsis as an outcome variable, and perform LASSO regression on the expression of the iron death-related genes as a feature vector to obtain genes meeting preset conditions; wherein, the loss function of the LASSO regression is as follows: where y i is the clinical outcome of the i-th patient, which is a binary variable indicating whether the patient is septic or not, y i = 1 indicates sepsis, and y i = 0 indicates non-sepsis; is the prediction value of the i-th patient, β is the vector of regression coefficients, n is the number of patients, λ is the regularization parameter, and ||β||1is the L1-norm of β. 5.The training method of a respiratory infection disease progression and prognosis prediction model according to claim 4, characterized in that, The step 5 comprises the following steps: Take whether a patient is sepsis as a binary outcome variable, and take the expression of the screened genes as a prediction variable, and establish an RTI clinical outcome prediction model through logistic regression, which comprises the following steps: Take whether a patient is sepsis as a binary outcome variable, and take the expression of the screened genes as a prediction variable, and establish an RTI clinical outcome prediction model through logistic regression; wherein, the logistic regression algorithm is as follows: wherein is a logit transformation that maps the probability P to the real number domain, β0is an intercept term, β1is the first regression coefficient, Y is the sepsis outcome variable, and X1is the PGD expression level in the peripheral blood of the patient with acute respiratory tract infection. 6.The training method of a respiratory infection disease progression and prognosis prediction model according to claim 1, characterized in that, logit(P(Y=1))=β0+β1X1 In step 5, the following steps are further included: Step 5.1: Evaluate the RTI clinical outcome prediction model to obtain a plurality of evaluation indexes; Step 5.2: Normalize each evaluation index to obtain normalized evaluation indexes; Step 5.3: Determine a weighting coefficient according to the proportion of each evaluation index in all evaluation indexes; Step 5.4: Weighted sum the normalized evaluation indexes to obtain a comprehensive evaluation index; 7.The training method of a respiratory infection disease progression and prognosis prediction model according to claim 6, characterized in that, Step 5.5: When the comprehensive evaluation index is in a preset range, the RTI clinical outcome prediction model is trained. In step 5.3, the formula is as follows: Determine a weighting coefficient according to the proportion of each evaluation index in all evaluation indexes; wherein P i represents the proportion of the i-th evaluation index, x i represents the normalized value of the i-th evaluation index, m represents the number of evaluation indexes, w i represents the weight of the i-th index.

8. An electronic device comprising a bus, a transceiver, a memory, a processor, and a computer program stored on the memory and executable on the processor, the transceiver, the memory and the processor being connected by the bus, characterized in that, The computer program, which is executed by the processor, implements the steps of a training method of a respiratory tract infection disease progression and prognosis prediction model as claimed in any one of claims 1-7.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program, which is executed by the processor, implements the steps of a training method of a respiratory tract infection disease progression and prognosis prediction model as claimed in any one of claims 1-7.

Images