Early risk prediction model for ischemia-reperfusion acute kidney injury based on ferroptosis-related gene and construction method and application thereof

Abstract

The invention discloses an ischemia-reperfusion acute kidney injury early risk prediction model based on ferroptosis related genes and a construction method and application thereof. The method comprises the following steps: screening five core genes: SLC39A14, TFRC, HMOX1, TXNRD1 and GCLM by using differential expression analysis in combination with a random forest algorithm and a support vector machine recursive feature elimination method; constructing a column graph model through multi-factor Logistic regression, mapping the expression levels of the five core genes into prediction scores, and generating an AKI occurrence risk score; according to the invention, the column diagram model is combined with ferroptosis and immune characteristics, and is more in line with the pathogenesis of AKI; the risk of a patient can be directly quantified, and clinical application is facilitated; through internal verification, the model has relatively high accuracy and clinical practical value.

Description

Technical Field

[0001] This invention relates to an early risk prediction model for ischemia-reperfusion acute kidney injury (IR-AKI) based on ferroptosis-related genes, its construction method, and its application, belonging to the field of biomedical detection technology. Background Technology

[0002] In current technologies, the diagnosis of acute kidney injury due to renal ischemia-reperfusion mainly relies on changes in serum creatinine and urine output. However, these indicators often lack sensitivity and specificity in the early stages, making timely intervention difficult. Although some predictive models based on biomarkers exist, most are single-gene or multi-gene combination models, lacking integrated analysis of ferroptosis and immune-related factors. They cannot accurately reflect changes in proximal tubular cell function and inflammatory status, nor do they fully utilize high-throughput transcriptome data and machine learning methods for systematic screening and validation. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by providing an early prediction model for IR-AKI that combines ferroptosis-related genes and immune characteristics. By integrating multi-omics data and machine learning algorithms, this model improves the accuracy and reliability of early diagnosis of ischemia-reperfusion acute kidney injury, thereby providing a basis for clinical decision-making and can be used in scientific research to evaluate IR-AKI-related targets or intervention effects.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a method for constructing an early risk prediction model for ischemia-reperfusion acute kidney injury based on ferroptosis-related genes, comprising the following steps: Step 1: First, collect renal tissue transcriptome data from IR-AKI patients and healthy controls according to the inclusion and exclusion criteria; Step 2: Use differential expression analysis to screen for differentially expressed genes related to ferroptosis, and combine machine learning algorithms to screen and identify five core genes for model construction: SLC39A14, TFRC, HMOX1, TXNRD1, and GCLM. Step 3: Based on the five core genes selected in Step 2, construct a nomogram model using multivariate logistic regression; Step 4: Model performance evaluation and validation; The "early period" refers to the window of 0–1 hour after renal ischemia-reperfusion.

[0005] Preferably, the specific steps of screening using machine learning algorithms in step 2 include: ranking the candidate genes screened by differential expression analysis by importance using the random forest algorithm, determining the top ten ferroptosis-related genes based on the importance ranking, and then further screening five key genes by the support vector machine recursive feature elimination method. The Venn diagram shows the intersection of the screening results of differential expression analysis, Random Forest, and SVM-RFE, and finally determines five core genes.

[0006] Preferably, in step 4, the model prediction performance is evaluated using AUC, calibration curve, and DCA curve.

[0007] Preferably, in step 4, the training group and the validation group are randomly split, and the model is validated using the Bootstrap method.

[0008] The present invention also provides a risk prediction model obtained by the above construction method.

[0009] The present invention also provides the application of the above-mentioned risk prediction model in the preparation of products for early risk prediction of ischemia-reperfusion acute kidney injury.

[0010] This invention also provides an early risk prediction system for ischemia-reperfusion acute kidney injury, comprising: The data acquisition module is used to acquire transcriptome data of the subject's kidney tissue or gene expression data of five genes: SLC39A14, TFRC, HMOX1, TXNRD1 and GCLM. A data parsing module, connected to the acquisition module, is used to input the renal tissue transcriptome data or gene expression data acquired by the data acquisition module into the pre-constructed nomogram model and calculate the total risk score. Based on the total risk score, a predicted value P(AKI) of the early risk probability of ischemia-reperfusion acute kidney injury is calculated. The nomogram model is the risk prediction model obtained by the above construction method. And an output module, connected to the parsing module, for outputting a predicted value of the probability of early risk of acute kidney injury in the subject.

[0011] Preferably, the risk scoring formula corresponding to the nodal chart model is: RiskScore = 2.019 + 0.818 × SLC39A14 + 2.072 × TFRC + 1.278 × HMOX1 + 0.104 × TXNRD1 + 0.328 × GCLM, where each gene represents the gene expression level after normalizing the expression matrix using the limma package in R. The formula for calculating P(AKI) is as follows:

[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: Compared with traditional single-index prediction methods, the model of this invention has the following advantages: 1) Combining ferroptosis and immune characteristics, it is more consistent with the pathogenesis of IR-AKI; 2) The core genes selected by using high-throughput data and machine learning methods are robust and reliable; 3) The nomogram can directly quantify patient risk, which is convenient for clinical application; 4) The model has been internally validated by the Bootstrap method and has high accuracy and clinical practical value. Attached Figure Description

[0013] Figure 1 This study describes the construction and validation of a diagnostic model for ferroptosis-related ischemia-reperfusion acute kidney injury. The model is presented as follows: a) Performance curves of candidate genes ranked by importance using a Random Forest algorithm; b) Ten ferroptosis-related genes identified based on the importance ranking in a); c) Expression validation results of five key genes further screened using Support Vector Machine Recursive Feature Elimination (SVM-RFE); d) Venn plot showing the intersection of differential expression analysis, Random Forest, and SVM-RFE results; e) Forest plot of multivariate logistic regression analysis, displaying the odds ratio (OR) and 95% confidence interval for the five ferroptosis-related genes, showing that only TFRC was statistically significant; f) Model performance evaluation using ROC curves; g) Model robustness validation using Bootstrap; h) Model robustness analysis using calibration curves; i) Nomogram constructed based on the five ferroptosis-related genes for predicting individual AKI risk; j) Clinical net benefit of the decision curve comparison analysis (DCA) model and a single gene. Detailed Implementation

[0014] To make the present invention more apparent and understandable, preferred embodiments are described in detail below with reference to the accompanying drawings.

[0015] Unless otherwise specified, the test methods or experimental methods described in the following examples are conventional methods; unless otherwise specified, the reagents and materials are obtained from conventional commercial sources or prepared by conventional methods.

[0016] Example A method for constructing an early risk prediction model for IR-AKI based on ferroptosis-related genes includes the following steps: Step 1: First, we collected renal tissue transcriptome data from IR-AKI patients and healthy controls. The data sources, processing methods, and grouping are as follows: 1) Data source: This embodiment is based on the publicly available transcriptome dataset GSE43974 from the GEO database of the National Center for Biotechnology Information (NCBI).

[0017] 2) Data processing: i) Data download and initial processing Use GEOquery::getGEO("GSE43974") to download the expression matrix and sample annotations (pdata), and remove unnecessary donor types based on the sample annotations (excluding donor type = 2), keeping only the control and IR-AKI group samples for subsequent analysis; ii) Normalization To eliminate technical differences between different chips, the original expression matrix was normalized using limma::normalizeBetweenArrays() (this function implements normalization based on the quantile method), and the distribution changes before and after normalization were examined using a boxplot to confirm the effect (code: exprSet=normalizeBetweenArrays(exprSet)).

[0018] 3) Data inclusion criteria: i) Kidney tissue biopsy from the early stage of intraoperative reperfusion or from a healthy donor; ii) Chip platform is consistent (GPL6244); iii) Complete clinical information; 4) Data exclusion criteria: i) Samples from cardiac death donors (DCD, donor type 2); ii) Early tissue samples without reperfusion; 5) Following the inclusion and exclusion criteria described above, and based on the time point field (characteristics_ch1.2) in the sample annotations, samples taken 60 min after reperfusion (timepoint biopsy: 1) were selected as the IR time point for subsequent differential and enrichment analyses. The final subset exprSet2 used for analysis was obtained (final sample size: 138; AKI = 101, control = 37), meaning a total of 138 samples were ultimately included. IR-AKI group: 101 cases (brain-dead donors, reperfused with kidney tissue 60 min postoperatively) Control group: 37 cases (kidney tissue from healthy living donors).

[0019] In the GSE43974 dataset, renal iron homeostasis and injury signals can be detected by sampling 60 minutes after reperfusion, which is clinically recognized as the earliest predictable stage of AKI. Therefore, the "early stage of acute kidney injury" described in this invention refers to the hyperacute window of 0–1 hour after reperfusion, corresponding to the stage before the molecular-level damage caused by acute ischemia-reperfusion has caused significant structural changes.

[0020] Step 2: Differential expression analysis (DESeq2 or limma package) is used to screen for differentially expressed genes related to ferroptosis, and GSVA is used to calculate the ferroptosis signal enrichment score for each sample to confirm that the ferroptosis pathway is activated in AKI, providing theoretical support for the differential analysis. The selected candidate genes are then input into a Random Forest algorithm for importance ranking, and its performance curve is shown in the figure. Figure 1 As shown in a, the top ten ferroptosis-related genes were determined based on their importance. Figure 1 a, 1b). Subsequently, five key genes were further screened using the Support Vector Machine Recursive Feature Elimination (SVM-RFE) method, and their expression verification results are as follows. Figure 1 As shown in Figure c, the Venn diagram illustrates the intersection of differential expression analysis, Random Forest, and SVM-RFE screening results, ultimately identifying five core genes: SLC39A14, TFRC, HMOX1, TXNRD1, and GCLM. Figure 1 d). Forest plots generated by multivariate logistic regression analysis were used to show the odds ratios (ORs) and 95% confidence intervals of the five ferroptosis-related genes. The results showed that only TFRC was statistically significant (1e).

[0021] Step 3: Construct a nomogram (1i) using multivariate logistic regression to map the expression levels of the five core genes to predicted scores, generating an IR-AKI risk score. The scoring formula is as follows: RiskScore=2.019+0.818×SLC39A14+2.072×TFRC+1.278×HMOX1+0.104×TXNRD1+0.328×GCLM; The gene expression values ​​are derived from the standardized expression matrix (normalized using the limma package in R). The higher the Risk Score, the greater the IR-AKI risk.

[0022] The probability of IR-AKI occurrence output by the model is calculated using the Logistic function:

[0023] Step 4: Model Performance Evaluation To avoid overfitting and evaluate the model's generalization ability, the Bootstrap method (R = 1000 repetitions) was used for internal validation. The specific method is as follows: 1) Construct a Logistic Regression Model using the entire sample; 2) In each Bootstrap resampling: i) Randomly draw a resampled set with replacement, with the same sample size as the original sample, for model fitting; ii) Apply the model's returned predictions to the complete original dataset; iii) Calculate ROC-AUC, sensitivity, and specificity; 3) Calculate the mean and 95% confidence intervals for AUC, sensitivity, and specificity; 4) Plot the ROC curve and the model calibration curve to evaluate the performance.

[0024] Model performance was evaluated using ROC curves, showing a high AUC value of 0.908 ( Figure 1 f), Bootstrap validation and calibration curve analysis further confirmed the model's robustness. Figure 1 g, 1h). Nodal plot prediction performance is significantly better than that of a single gene ( Figure 1 i) Decision curve analysis (DCA) shows that within a certain threshold range, nomograms can provide a higher net clinical benefit (i). Figure 1 j).

[0025] In this invention, RNA can be extracted from a patient's blood or kidney tissue sample, and the expression levels of core genes can be measured. By inputting the gene expression values ​​into a nomogram model or calculating a risk score, the probability of the patient developing acute kidney injury (AKI) can be obtained. Based on the predicted risk level, clinicians can be guided to implement early interventions, such as enhanced fluid management, pharmacological intervention, or close monitoring of kidney function. Simultaneously, this model can be used in clinical trials or drug development to assess the impact of interventions on AKI risk, helping to screen potential targets or validate treatment efficacy.

[0026] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any form or substance. It should be noted that those skilled in the art can make several improvements and additions without departing from the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

Claims

1. A method for constructing an early risk prediction model for ischemia-reperfusion acute kidney injury based on ferroptosis-related genes, characterized in that, Comprising the following steps: Step 1, first collect renal tissue transcriptome data from IR-AKI patients and healthy controls according to the standard of environmental protection and carry out data preprocessing; Step 2, use differential expression analysis to screen iron death related differential genes, and combine machine learning algorithm to screen and determine five core genes for model construction: SLC39A14, TFRC, HMOX1, TXNRD1 and GCLM; Step 3, based on the five core genes screened in step 2, a nomogram model is constructed by multi-factor logistic regression; Step 4, model performance evaluation and verification; The early stage refers to the window period within 0-1 hour after renal ischemia-reperfusion.

2. The construction method according to claim 1, characterized in that, The specific steps of screening in step 2 combined with machine learning algorithm include: sorting the candidate genes screened by differential expression analysis through random forest algorithm according to importance, and determining the top ten iron death related genes according to importance ranking, and then further screening five key genes through support vector machine recursive feature elimination method, Venn diagram shows the intersection of differential expression analysis, Random Forest and SVM-RFE screening results, and finally determines five core genes.

3. The construction method of claim 1, wherein, The model prediction performance is evaluated by AUC, calibration curve and DCA curve in step 4.

4. The construction method of claim 1, wherein, The training group and the verification group are randomly divided and the Bootstrap method is used for model verification in step 4.

5. The risk prediction model obtained by the construction method according to any one of claims 1-4.

6. The use of the risk prediction model of claim 5 in the preparation of a product for early risk prediction of ischemia-reperfusion acute kidney injury.

7. An early risk prediction system for ischemia-reperfusion acute kidney injury, characterized by, Comprising: a data acquisition module for acquiring renal tissue transcriptome data or gene expression data of five genes: SLC39A14, TFRC, HMOX1, TXNRD1 and GCLM; a data analysis module connected with the acquisition module, for inputting the renal tissue transcriptome data or gene expression data acquired by the data acquisition module into the nomogram model which has been constructed and calculating the total risk score, and calculating the prediction value P(AKI) of the probability of early risk of ischemia-reperfusion acute kidney injury according to the total risk score; wherein the nomogram model is the risk prediction model obtained by the construction method according to any one of claims 1-4; and an output module connected with the analysis module, for outputting the prediction value of the probability of early risk of acute kidney injury in the subject.

8. The risk prediction system of claim 7, wherein, The risk score formula corresponding to the nomogram model is: RiskScore=2.019+0.818×SLC39A14+2.072×TFRC+1.278×HMOX1+0.104×TXNRD1+0.328×GCLM, wherein each gene represents the gene expression after normalization of the expression matrix using the limma package in R language; The calculation formula of P(AKI) is as follows: 。

Images