Model for predicting AMI occurrence risk and kit for predicting AMI occurrence risk

Abstract

The invention discloses a model for predicting an AMI occurrence risk and a kit for predicting the AMI occurrence risk. According to the method, the high-risk CVD patient is identified before the AMI completely attacks, and the purpose of earlier-stage monitoring is achieved. Meanwhile, the combination of CVD and AMI and pure CVD are selected as comparison objects, and wide inflammation background interference caused by basic cardiovascular diseases is effectively eliminated, so that the recognition specificity of the AMI-related microRNA is remarkably improved.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of biomedicine and artificial intelligence, and in particular to a model for predicting the risk of acute myocardial infarction (AMI) and a reagent kit for predicting the risk of AMI. Background Technology

[0002] Acute myocardial infarction (AMI) is a clinical syndrome caused by the formation of a thrombus following the rupture or erosion of atherosclerotic plaques in the coronary arteries, leading to persistent ischemia, hypoxia, and necrosis of myocardial tissue. AMI is the most urgent and deadliest type of cardiovascular disease and is one of the leading causes of death and disability worldwide. According to statistics from the World Health Organization (WHO) and epidemiological data from various countries, acute myocardial infarction (AMI) causes millions of deaths annually, and the incidence rate is increasing year by year. Early identification and timely intervention are key to reducing AMI mortality and improving prognosis. Currently, the clinical diagnosis of AMI mainly relies on the following aspects: Clinical manifestations include persistent chest pain, sweating, and difficulty breathing, but some patients (such as the elderly and women) may present with atypical symptoms. Electrocardiogram (ECG) and Holter monitoring: ST segment elevation. Elevated serum biomarkers, such as troponin (cTnI, cTnT) and other biomarkers. Imaging techniques: such as coronary angiography, echocardiography, etc. In conclusion, the severity of acute myocardial infarction (AMI) and its sudden onset necessitate the development of earlier and more sensitive biological prediction methods to enhance the ability to proactively identify and intervene in high-risk populations. While the methods described above are important in the diagnosis of acute myocardial infarction (AMI), they share a common limitation: they can only detect relevant changes after the myocardium has already been damaged, lacking the ability to provide early warning for high-risk individuals. For example, elevated troponin levels usually occur several hours after a myocardial infarction; electrocardiogram changes are not obvious in some atypical patients; and imaging examinations are expensive and difficult to apply to widespread screening. In addition, some studies are currently attempting to develop predictive tools by comparing biomarker differences between AMI patients and healthy individuals, but this approach ignores the fact that in actual clinical practice, AMI often occurs in people with pre-existing cardiovascular and cerebrovascular diseases (CVD).

[0003] Therefore, existing technologies have shortcomings and need to be improved. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a model for predicting the risk of AMI and a kit for predicting the risk of AMI, so as to solve the problems mentioned in the background art.

[0005] The technical solution of this invention is as follows: A model for predicting the risk of acute myocardial infarction (AMI) is provided, established using the following steps: S1: The GSE123342 dataset, publicly available in the GEO database, was selected. This dataset contains peripheral blood microRNA expression data of AMI patients with CVD and CVD patients alone. Missing samples were removed, and log2 transformation and expression value normalization were performed to ensure comparability between samples. S2: Two-group comparisons were used to perform independent samples t-tests on each microRNA probe to calculate its logFC (log2 Fold Change) and original P-value (original p-value); then, a threshold P-value < 0.05 was used to initially screen probes with suggestive significance. To visually represent the differential results, a volcano plot based on logFC and -log10 (p-value) was created. Probes were categorized into three types: upregulated, downregulated, and not significant (Not Sig); red indicates upregulation, blue indicates downregulation, and gray indicates not significant; the horizontal axis represents logFC, and the vertical axis represents -log10 (p-value). The gene names of the five most significant probes were labeled in the plot.

[0006] S3: Select the top 20 microRNAs with the most significant p-values ​​in the differential analysis, extract their expression values ​​in all samples, and standardize them; divide the standardized data into training and test sets, with the training set comprising 60%-80% and the test set comprising 40%-20%. Then, a logistic regression model is built, the training set is used to fit the logistic regression model, and the performance of the logistic regression model is evaluated on the test set. After the logistic regression model was fitted, the regression coefficient of each microRNA was further extracted to assess its importance in the logistic regression model; the positive or negative sign of the coefficient represents its directional effect on the risk of AMI, and the absolute value represents the strength of its influence. S4: Perform feature compression and model optimization on the established logistic regression model; specifically, reduce the number of variables through stepwise feature compression while maintaining model performance, as follows: A1: Based on the established logistic regression model, microRNAs were ranked according to the importance of the regression coefficients; A2: Incrementally add the first N microRNA variables in the sorting and establish logistic regression models for each; for example, the first, second, third, fourth, fifth, sixth, seventh, and eighth, etc., and establish logistic regression models for each. A3: Record the AUC and accuracy trend for each set of variables; A4: Observe the results and determine that using the data from the first M microRNAs can achieve the effect of the logistic regression model in step S3, thus optimizing the logistic regression model. S5: To enhance nonlinear expression capabilities, a random forest model was used for modeling, with the data from the first M microRNAs selected in step S4 used for training and evaluation.

[0007] S6: After the random forest model is trained and evaluated, in order to enable model reuse and deployment, the final random forest model and its corresponding normalizer are serialized and saved.

[0008] In step S3, the expression values ​​are standardized by scaling the features to make them have the same mean and variance, thereby improving the stability of the model.

[0009] In step S5, the data of the first M microRNAs are divided into a training set and a test set, with the training set accounting for 60%-80% and the test set accounting for 40%-20%.

[0010] Use the joblib tool to export the trained random forest model object (rf_model) and the StandardScaler object (scaler) used for data standardization as random_forest_model.pkl and scaler.pkl files, respectively.

[0011] When calling this random forest model, the model and normalizer are loaded separately using joblib.load().

[0012] The M is 8, and the 8 microRNAs are: hsa-mir-4520-2, hsa-mir-2052, hsa-mir-505-3p, hsa-mir-4457, hsa-mir-27A-3p, hsa-mir-5047, hsa-mir-3651, and hsa-mir-635.

[0013] This invention provides a kit for predicting the risk of acute myocardial infarction (AMI), comprising the detection reagents hsa-mir-4520-2, hsa-mir-2052, hsa-mir-505-3p, hsa-mir-4457, hsa-mir-27A-3p, hsa-mir-5047, hsa-mir-3651, and hsa-mir-635.

[0014] The kit for predicting the risk of AMI is used to predict the risk of AMI, and its application method includes the following steps: B1: MicroRNA Sequence Acquisition; To achieve high sensitivity and specificity in peripheral blood microRNA detection, based on the human mature microRNA information registered in the miRBase database, the 3p / mature sequences of 8 target microRNAs were acquired: hsa-mir-4520-2, hsa-mir-2052, hsa-mir-505-3p, hsa-mir-4457, hsa-mir-27A-3p, hsa-mir-5047, hsa-mir-3651, and hsa-mir-635. A detection system for these 8 target microRNAs was established. The detection system was designed based on the TaqMan probe-based qPCR principle, balancing specificity and detection sensitivity. B2: Stem-loop RT primer and qPCR amplification primer design; cDNA was constructed using the mature stem-loop primer method, and qPCR amplification was performed using universal reverse primers and specific forward primers; each microRNA probe sequence carries a FAM fluorescent group and a BHQ1 quencher group to achieve highly specific fluorescence detection; the primer and probe sequences are shown in the table below:

[0015] B3: Sample collection and extraction; specifically including: B31: After whole blood is drawn, serum or plasma is separated according to standard centrifugation method, and total RNA is extracted using a commercially available RNA extraction kit; it is stored in a timely manner to ensure that the RNA purity meets the standard and there is no obvious degradation; U6 snRNA is used as the standard internal control RNA, and eight candidate microRNAs that have been validated are selected as positive controls to ensure the effectiveness of the amplification reaction and the accuracy of interpretation. B32: Reverse transcription reaction; the extracted total RNA is reverse transcribed to generate cDNA of microRNA; each target microRNA uses a specially designed stem-loop RT primer, with the 3' end complementary to the end of the microRNA sequence and containing a fixed universal tail sequence to ensure consistent pairing in subsequent amplification; B33: RT-qPCR amplification; After reverse transcription, the obtained cDNA is used for qRT-PCR amplification and detection; Each microRNA is detected using a specific forward primer, a universal reverse primer, and a specific TaqMan fluorescent probe (5'-FAM-sequence-BHQ1-3') that is complementary to the target sequence midline, in conjunction with a TaqMan fluorescent probe-specific reaction system; B34: After amplification, output the Ct values ​​of each target microRNA and internal reference RNA for subsequent data standardization and risk assessment; B4: Data Transformation and Input; The Ct (Cycle Threshold) value obtained by qRT-PCR detection reflects the relative expression level of each microRNA; Since the constructed prediction model is based on the GEO chip expression matrix, the Ct values ​​obtained from actual detection need to be converted into equivalent expression values ​​using Formula 1. log2(expression) = (29.67) Ct) / 0.94 (Formula 1), Among them, 29.67 and 0.94 are the intercept and slope obtained by regression fitting of the standard curve, respectively.

[0016] The transformed log2(expression) value is then input into the final random forest model obtained above for analysis.

[0017] The reaction system for RT-qPCR amplification is as follows:

[0018] The amplification program was fixed as follows: initial denaturation at 95℃ for 5 minutes, followed by 40-45 cycles. Each cycle included denaturation at 95℃ for 15 seconds and annealing / extension at 60℃ for 60 seconds. Fluorescence signals were recorded during the cycle, and the amplification curve of the target sequence was detected in real time.

[0019] The reverse transcription system contains reverse transcriptase, dNTPs, RNase, inhibitors, and buffer. The primer binding, reverse transcription, and enzyme inactivation are completed by setting the required temperature program to generate a specific cDNA template.

[0020] The present invention has the following technical advantages 1. Novel microRNA Biomarker Combination: This invention proposes a combination of eight microRNAs with specific expression patterns as biomarkers to identify the risk of acute myocardial infarction (AMI) in CVD patients. These microRNAs exhibit stable abnormal expression before the onset of AMI, demonstrating high sensitivity and good biological stability, and are an important supplement to traditional troponin-based biomarkers. 2. Applicable to early risk identification: Traditional AMI diagnostic methods mostly rely on physiological or biochemical abnormalities that only appear after the event, such as elevated troponin or ST segment elevation, which has a significant lag. This technology, however, through abnormal microRNA expression, can identify AMI before typical clinical symptoms appear, facilitating prospective monitoring and proactive intervention in high-risk populations. 3. The detection method is minimally invasive and easy to promote: microRNA is extracted from peripheral blood samples and quantitatively analyzed by digital PCR platform using the experimental design of this patent. No biopsy or large imaging equipment is required, making it feasible for widespread application in primary hospitals and even physical examination institutions. 4. Innovative Clinical Modeling Strategy: Unlike traditional research methods that use healthy individuals as the control group, this invention adopts a comparative strategy of "CVD patients vs. CVD patients with AMI". This strategy can remove the interference of the underlying disease background, accurately identify the specific signals of AMI events, and make the prediction model more consistent with real clinical scenarios, thus making it more practical and reliable. 5. Machine Learning-Based Model Construction and Iterative Optimization: The random forest algorithm is used for feature selection and model training, ultimately establishing a high-efficiency classifier with an accuracy exceeding 85% and an AUC exceeding 0.80. The model supports continuous updates with subsequent data, possesses good iterative properties, and can be continuously optimized with the accumulation of samples, thereby enhancing its clinical application value.

[0021] By adopting the above scheme, the present invention provides a model for predicting the risk of AMI and a kit for predicting the risk of AMI, which has the following technical effects: 1. This invention, based on the combination of differentially expressed microRNA features, accurately identifies acute myocardial infarction (AMI) events in patients with coronary heart disease (CVD), significantly improving the sensitivity and specificity of diagnosis and contributing to the optimization of early intervention and treatment decisions. 2. Compared with traditional single biomarker detection methods, this technology uses multi-microRNA feature joint modeling, achieving higher model accuracy (0.862) and AUC (0.838), providing clinicians with more reliable predictive basis. 3. The established random forest classification model only requires the expression levels of 8 key microRNAs to complete the diagnosis, balancing detection efficiency and clinical applicability, and has good translational potential. In addition, the accompanying experimental content was designed to achieve rapid quantitative detection and standardized result output through a digital PCR platform. 4. The model has been standardized and saved, has plug-and-play capability, and supports automatic iterative optimization through new data to ensure its continuous and efficient performance in multi-center and multi-population environments. 5. This technology provides a repeatable, updatable, and portable intelligent diagnostic framework that can be further extended to other cardiovascular diseases or multi-omics joint analysis scenarios, and has broad market prospects and scientific research application value. Attached Figure Description

[0022] Figure 1 Volcano diagram of differential expression of microRNA probes.

[0023] Figure 2 The accuracy and area under the curve of using 20 microRNAs in a logistic regression model to identify AMI patients were calculated.

[0024] Figure 3 This is a graph showing the relationship between the number of microRNAs used for modeling and the performance of the logistic regression model.

[0025] Figure 4 This is a graph showing the correspondence between the number of microRNAs used for modeling and the performance of the random forest model. Detailed Implementation

[0026] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0027] The pathogenesis of acute myocardial infarction (AMI) mainly involves the rupture of coronary atherosclerotic plaques, platelet aggregation, and thrombus formation, ultimately leading to ischemic necrosis of myocardial tissue. Recent studies have found that some microRNAs exhibit stable expression changes in peripheral blood before the onset of AMI, serving as potential precursor biomarkers. This technical solution is based on molecular-level abnormalities in the early pathological process of AMI. It selects eight microRNAs that are specifically expressed in CVD patients during the onset of AMI. Through high-throughput screening, statistical analysis and machine learning training, a combined biomarker model is constructed to achieve highly sensitive identification of AMI. Compared to traditional methods that rely on clinical manifestations and cTn elevation after an acute myocardial infarction (AMI), this invention aims to identify "high-risk CVD patients" before AMI fully develops, achieving earlier monitoring. Furthermore, the study design uses "CVD combined with AMI" and "CVD alone" as comparison subjects, effectively eliminating interference from the widespread inflammatory background caused by underlying cardiovascular disease, thereby significantly improving the specificity of AMI-related microRNA recognition.

[0028] I. Data Collection and Preprocessing The GSE123342 dataset, publicly available in the GEO database, was selected. It contains peripheral blood microRNA expression data of AMI patients (with CVD) and CVD patients. Missing samples were removed, and log2 transformation and expression value normalization were performed. This data processing method is based on Hanyu, Y., Z. Xiaoyong, and W. Lei, Review of DataNormalization Methods. Computer Engineering and Applications, 2023. 59(3): p.13-22, to ensure comparability between samples. II. Difference Screening Two-sample t-tests were performed on each microRNA probe using a two-group comparison approach to calculate its log2 Fold Change (logFC) and original p-value. This data processing method is referenced from Rosati, D., et al., Differential gene expression analysis pipelines and bioinformatic tools for the identification of specific biomarkers: A review. Comput StructBiotechnol J, 2024. 23: p. 1154-1168. Then, a threshold p-value < 0.05 was used to initially screen probes with suggestive significance. Please refer to [link to relevant documentation]. Figure 1 To visually represent the differences, a volcano plot based on logFC and -log10 (p-value) was created. Probes were categorized into three types: upregulated, downregulated, and not significant (Not Sig); red indicates upregulation, blue indicates downregulation, and gray indicates not significant; the horizontal axis represents logFC, and the vertical axis represents -log10 (p-value). Figure 1 The gene names of the top five most prominent probes are labeled. The volcano plot reveals candidate microRNAs that are specifically upregulated or downregulated in AMI relative to CVD status, providing a key feature set for subsequent modeling.

[0029] III. microRNA Selection and Feature Construction 3.1 Preliminary Modeling and Evaluation (1) Preliminary modeling and evaluation First, the top 20 microRNAs with the most significant p-values ​​in the differential analysis (such as MIR2052, MIR5047, MIR27A, etc.) were selected, and their expression values ​​in all samples were extracted and standardized. Then, a logistic regression model was used to fit the model on 70% of the training set and to evaluate the model performance on 30% of the test set. Please see Figure 2 The model results show that when using these 20 microRNA variables, the model exhibits good discriminative power on the test set: • The accuracy rate is 0.862. • The area under the curve (AUC) is 0.835. Among them, accuracy represents the proportion of correctly predicted samples, and AUC reflects the model's ability to distinguish between positive (AMI) and negative (CVD) samples. The closer the AUC is to 1, the better the performance. See Jin, H. and CX Ling, Using AUC and accuracy in evaluating learning algorithms. IEEE Transactions on Knowledge and Data Engineering, 2005. 17(3): p. 299-310. The results at this stage demonstrate that microRNA has significant information content in AMI discrimination. After model fitting, regression coefficients for each microRNA were extracted to assess its importance in the model. The sign of the coefficient represents its directional effect on AMI risk, while the absolute value represents the strength of its influence. The following are the top 10 microRNA features sorted by absolute coefficient and their model weights:

[0030] (2) Feature compression and model optimization Considering the high cost of testing 20 microRNAs at once in clinical trials, we attempted to reduce the number of variables through stepwise feature compression while maintaining model performance: • Based on the logistic regression model, microRNAs were ranked according to the importance of the regression coefficients; • Incrementally add the first 1–10 microRNA variables and build models accordingly; • Record the AUC and accuracy trends for each set of variables; • Please see Figure 3 The results showed that using the first 8 microRNAs could achieve model performance similar to that of 20 microRNAs, and had a high cost-effectiveness for clinical application.

[0031] Furthermore, to enhance the model's nonlinear expressive power, a random forest model is employed for modeling. Please refer to [link / reference]. Figure 4 Under the same 8 microRNA features, the model achieved the following performance on the test set: Random Forest accuracy: 0.862, AUC: 0.838.

[0032] This demonstrates that the constructed random forest model can still achieve a high level of AMI discrimination even with only 8 microRNAs being detected, providing a good feasibility for its application in actual clinical practice. The relative weights of each microRNA were also derived for subsequent model iterations.

[0033]

[0034] (3) Model Derivation and Iterative Optimization After model training and evaluation, to enable model reuse and deployment, this study serialized and saved the final random forest model and its corresponding normalizer. Specifically, the joblib tool was used to export the trained random forest model object (rf_model) and the StandardScaler object (scaler) used for data normalization as random_forest_model.pkl and scaler.pkl files, respectively. This allows for direct loading of the model for prediction in subsequent clinical applications or diagnostic software development, as well as for iterating on the model using new data.

[0035] In the future, when calling this model, simply load the model and normalizer using joblib.load(). The specific code is as follows: import joblib # Loading Models and Normalizers model = joblib.load("saved_models / random_forest_model.pkl") scaler = joblib.load("saved_models / scaler.pkl") # Standardize the new samples X_new_scaled = scaler.transform(X_new) # Make predictions prediction = model.predict(X_new_scaled) probability = model.predict_proba(X_new_scaled)[:, 1].

[0036] Furthermore, the model possesses iterative optimization capabilities, providing a solid foundation for future continuous development. As the clinical sample size accumulates, researchers can incorporate newly added AMI and CVD patient expression data into the existing training set, continuously improving the model's generalization ability and predictive accuracy by retraining the model or employing incremental learning strategies. This iterative optimization mechanism not only adapts to biological differences arising from different populations, regions, or testing platforms, but also captures novel microRNA biomarkers emerging with medical advancements, thus ensuring the model maintains its leading diagnostic performance. Furthermore, this strategy also provides a foundation for building a dynamically updated "AMI Intelligent Diagnostic System," enabling the model to continue to improve itself with new data after clinical deployment, achieving higher sensitivity and specificity, and meeting the needs of personalized medicine development.

[0037] IV. RT-qPCR Primer and Probe Design 4.1 MicroRNA Sequence Acquisition To achieve highly sensitive and specific detection of peripheral blood microRNAs, this study used the human mature microRNA information registered in the miRBase database (refer to Kozomara, A., M. Birgaoanu, and S. Griffiths-Jones, miRBase: from microRNA sequences to function. Nucleic Acids Res, 2019. 47(D1): p. D155-D162.) to obtain the 3p / mature sequences (hsa-mir-4520-2, hsa-mir-2052, hsa-mir-505-3p, hsa-mir-4457, hsa-mir-27A-3p, hsa-mir-5047, hsa-mir-3651, hsa-mir-635) of eight target microRNAs and established a detection system for these eight target microRNAs. All sequences were derived from authoritative annotation libraries to ensure standardization and reproducibility across laboratories and platforms. The detection system is designed based on the TaqMan probe-based qPCR principle, balancing specificity and detection sensitivity. 4.2 Design of stem-loop RT primers and qPCR amplification primers This project utilizes the established stem-loop RT primer method to construct cDNA, and then combines universal reverse primers with specific forward primers for qPCR amplification. Each microRNA probe sequence carries a FAM fluorescent group and a BHQ1 quencher group, achieving highly specific fluorescence detection. The primer and probe sequences are as follows:

[0038] 4.3 Experimental Design 4.3.1 Sample Collection and Extraction The required sample for the test is peripheral blood serum. After whole blood is drawn, serum or plasma is separated according to standard centrifugation method, and total RNA is extracted using a commercially available RNA extraction kit. It is stored in a timely manner to ensure that the RNA purity meets the standard and there is no obvious degradation. During the test, internal control RNA needs to be prepared (U6 snRNA is used as the standard internal control, refer to (Chen, C., et al., Real-time quantification of microRNAs by stem-loop RT-PCR. Nucleic Acids Res, 2005. 33(20): p. e179.)), and the validated target microRNA (eight candidate microRNAs screened in this protocol) is selected as a positive control to ensure the effectiveness of the amplification reaction and the accuracy of interpretation. 4.3.2 Reverse transcription reaction The extracted total RNA needs to be reverse transcribed to generate cDNA from microRNA. Each target microRNA uses a specially designed stem-loop RT primer, with its 3' end complementary to the end of the microRNA sequence and containing a fixed universal tail sequence to ensure consistent pairing during subsequent amplification. The reverse transcription system contains reverse transcriptase, dNTPs, RNase inhibitors, and buffer. The primer binding, reverse transcription, and enzyme inactivation are completed according to the specified temperature program to generate a specific cDNA template. The specific system and parameters are as follows:

[0039] 4.3.3 RT-qPCR Amplification After reverse transcription, the obtained cDNA was used for quantitative real-time PCR (qRT-PCR) amplification and detection. Each microRNA was detected using a matching specific forward primer, a universal reverse primer, and a specific TaqMan fluorescent probe (5'-FAM-sequence-BHQ1-3') complementary to the target sequence mid-segment, in accordance with the TaqMan fluorescent probe-specific reaction system. The reaction system and conditions are as follows:

[0040] The amplification program was fixed as follows: initial denaturation at 95℃ for 5 minutes, followed by 40-45 cycles, each cycle consisting of denaturation at 95℃ for 15 seconds and annealing / extension at 60℃ for 60 seconds, during which fluorescence signals were recorded, and the amplification curve of the target sequence was detected in real time. After amplification, the Ct values ​​of each target microRNA and internal control RNA were output for subsequent data standardization and risk assessment.

[0041] V. Data Transformation and Data Input The Ct (Cycle Threshold) values ​​obtained by qRT-PCR detection reflect the relative expression level of each microRNA. Since the constructed prediction model is based on the GEO microarray expression matrix (in units of Log2 expression level or Z-score), the actual Ct values ​​obtained need to be converted to equivalent expression levels using the following mathematical formula: log2(expression) = (29.67) Ct) / 0.94 (Formula 1), Wherein, 29.67 and 0.94 are the intercept and slope obtained by regression fitting of the standard curve, respectively, derived from the reference document: Li, Y., et al., Systematic identification and validation of the reference genes from 60 RNA-Seq libraries in the scallop Mizuhopectenyessoensis. BMC Genomics, 2019. 20(1): p. 288. The transformed log2 (expression) value was used to analyze the standardized input model.

[0042] This invention provides a model for predicting the risk of acute myocardial infarction (AMI) and a kit for predicting the risk of AMI, which have the following technical advantages: 1. This invention, based on the combination of differentially expressed microRNA features, accurately identifies acute myocardial infarction (AMI) events in patients with coronary heart disease (CVD), significantly improving the sensitivity and specificity of diagnosis and contributing to the optimization of early intervention and treatment decisions. 2. Compared with traditional single biomarker detection methods, this technology uses multi-microRNA feature joint modeling, achieving higher model accuracy (0.862) and AUC (0.838), providing clinicians with more reliable predictive basis. 3. The established random forest classification model only requires the expression levels of 8 key microRNAs to complete the diagnosis, balancing detection efficiency and clinical applicability, and has good translational potential. In addition, the accompanying experimental content was designed to achieve rapid quantitative detection and standardized result output through a digital PCR platform. 4. The model has been standardized and saved, has plug-and-play capability, and supports automatic iterative optimization through new data to ensure its continuous and efficient performance in multi-center and multi-population environments. 5. This technology provides a repeatable, updatable, and portable intelligent diagnostic framework that can be further extended to other cardiovascular diseases or multi-omics joint analysis scenarios, and has broad market prospects and scientific research application value.

[0043] This invention employs a “CVD vs CVD+AMI” control design to screen microRNA combinations, focusing on incremental expression signals in AMI events, which differs from traditional studies that use healthy individuals or CAD populations as references. 2. Based on machine learning optimization, an efficient combinatorial model consisting of only 8 microRNAs is used to predict AMI, balancing technical performance (AUC≈0.84) and detection cost. 3. The model supports dynamic updates and self-iteration, and its generalization ability and prediction accuracy can be continuously improved with the addition of clinical data, thus possessing high sustainability. 4. This invention designs a matching RT-qPCR detection system, including specific stem-loop RT primers, target forward primers, universal reverse primers, and TaqMan fluorescent probes labeled with FAM / BHQ1, forming a stable and highly specific standardized detection scheme that can be directly used for packaging product-grade micro-molecular diagnostic kits.

[0044] Terminology Explanation microRNA (miRNA): A class of non-coding RNA molecules about 22 nucleotides long, which are widely involved in gene expression regulation and can serve as potential biomarkers in a variety of diseases. AMI: Acute Myocardial Infarction, is an acute event caused by coronary artery obstruction leading to myocardial ischemia and necrosis. CVD: Cardiovascular Disease, including various clinical diseases such as coronary heart disease, hypertension, and heart failure. AUC: Area Under the Curve, an important metric for measuring the overall classification performance of a model; the closer it is to 1, the stronger its predictive ability. Random Forest: An ensemble learning classification algorithm with strong nonlinear modeling capabilities and anti-overfitting performance, widely used in clinical prediction modeling. Standardization: One of the data preprocessing steps, which improves model stability by scaling features to have the same mean and variance.

[0045] Differential expression analysis: Statistical methods are used to identify molecules, such as genes and microRNAs, that show significant differences in expression levels between the disease group and the control group. RT-qPCR (Real-Time Quantitative Polymerase Chain Reaction) is a molecular detection technique that uses fluorescence signals to monitor the product amplification process for quantitative RNA analysis. It is typically used to accurately quantify microRNA or mRNA expression levels. The experimental procedure involves two steps: RNA extraction and reverse transcription (RT), followed by real-time PCR amplification and detection. Reference gene: A relatively stable RNA molecule expressed in each sample, used as a calibration standard for qPCR to eliminate systematic errors.

[0046] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A model for predicting the risk of AMI (Acute Myocardial Infarction), characterized in that, The following steps are used to establish it, including: S1: The GSE123342 dataset, publicly available in the GEO database, was selected. This dataset contains peripheral blood microRNA expression data of AMI patients with CVD and CVD patients alone. Missing samples were removed, and log2 transformation and expression value normalization were performed to ensure comparability between samples. S2: Two-group comparisons were used to perform independent samples t-tests on each microRNA probe to calculate its logFC and original P-value; then, a threshold P-value < 0.05 was used to initially screen out probes with significant suggestive value. S3: Select the top few microRNAs with the most significant p-values ​​in the differential analysis, extract their expression values ​​in all samples, and perform standardization; divide the standardized data into training and test sets, with the training set accounting for 60%-80% and the test set accounting for 40%-20%. Then, a logistic regression model is built, the training set is used to fit the logistic regression model, and the performance of the logistic regression model is evaluated on the test set. After the logistic regression model was fitted, the regression coefficients of each microRNA were further extracted to assess its importance in the logistic regression model; the sign of the coefficient represents its directional effect on the risk of AMI, and the absolute value represents the strength of its influence. S4: Perform feature compression and model optimization on the established logistic regression model; specifically, reduce the number of variables by progressive feature compression while maintaining model performance, as follows: A1: Based on the established logistic regression model, microRNAs were ranked according to the importance of the regression coefficients; A2: Incrementally add the first N microRNA variables from the sorting and establish logistic regression models for each; A3: Record the AUC and accuracy trends for each set of variables; A4: Observe the results and determine whether the data from the first M microRNAs can achieve the effect of the logistic regression model in step S3, thereby optimizing the logistic regression model; S5: To enhance nonlinear expression capabilities, a random forest model was used for modeling, and the data used was the data of the first M microRNAs selected in step S4 for training and evaluation. S6: After the random forest model is trained and evaluated, in order to enable model reuse and deployment, the final random forest model and its corresponding normalizer are serialized and saved.

2. The model for predicting the risk of AMI according to claim 1, characterized in that, In step S3, the expression values ​​are standardized by scaling the features to make them have the same mean and variance, thereby improving the stability of the model.

3. The model for predicting the risk of AMI according to claim 1, characterized in that, In step S5, the data of the first M microRNAs are divided into a training set and a test set, with the training set accounting for 60%-80% and the test set accounting for 40%-20%.

4. The model for predicting the risk of AMI according to claim 1, characterized in that, In step S6, the joblib tool is used to export the trained random forest model object and the StandardScaler object used for data standardization as random_forest_model.pkl and scaler.pkl files, respectively; when calling the random forest model, the model and the standardizer are loaded respectively through joblib.load().

5. The model for predicting the risk of AMI according to claim 1, characterized in that, In step S2, a volcano plot based on logFC and -log10 was drawn to visually demonstrate the difference results.

6. The model for predicting the risk of AMI according to claim 1, characterized in that, The M is 8, and the 8 microRNAs are: hsa-mir-4520-2, hsa-mir-2052, hsa-mir-505-3p, hsa-mir-4457, hsa-mir-27A-3p, hsa-mir-5047, hsa-mir-3651, and hsa-mir-635.

7. A kit for predicting the risk of AMI, characterized in that, The detection reagents include hsa-mir-4520-2, hsa-mir-2052, hsa-mir-505-3p, hsa-mir-4457, hsa-mir-27A-3p, hsa-mir-5047, hsa-mir-3651, and hsa-mir-635.

8. A kit for predicting the risk of AMI according to claim 7, characterized in that, Its application method includes the following steps: B1: Establish a detection system for HSA-MIR-4520-2, HSA-MIR-2052, HSA-MIR-505-3p, HSA-MIR-4457, HSA-MIR-27A-3p, HSA-MIR-5047, HSA-MIR-3651, and HSA-MIR-635. This detection system is based on the TaqMan probe-based qPCR principle, balancing specificity and detection sensitivity. B2: Stem-loop RT primer and qPCR amplification primer design; cDNA was constructed using the mature stem-loop primer method, and qPCR amplification was performed using universal reverse primers and specific forward primers; each microRNA probe sequence carries a FAM fluorescent group and a BHQ1 quencher group to achieve highly specific fluorescence detection; the primer and probe sequences are shown in the table below: B3: Sample collection and extraction; specifically including: B31: After whole blood is drawn, serum or plasma is separated according to standard centrifugation method, and total RNA is extracted using a commercially available RNA extraction kit; it is stored in a timely manner to ensure that the RNA purity meets the standard and there is no obvious degradation; U6 snRNA is used as the standard internal control RNA, and eight candidate microRNAs that have been validated are selected as positive controls to ensure the effectiveness of the amplification reaction and the accuracy of interpretation. B32: Reverse transcription reaction; the extracted total RNA is reverse transcribed to generate cDNA of microRNA; each target microRNA uses a specially designed stem-loop RT primer, with the 3' end complementary to the end of the microRNA sequence and containing a fixed universal tail sequence to ensure consistent pairing in subsequent amplification; B33: RT-qPCR amplification; After reverse transcription, the obtained cDNA is used for qRT-PCR amplification and detection; Each microRNA is detected using a specific forward primer, a universal reverse primer, and a specific TaqMan fluorescent probe (5'-FAM-sequence-BHQ1-3') that is complementary to the target sequence midline, in conjunction with a TaqMan fluorescent probe-specific reaction system; B34: After amplification, output the Ct values ​​of each target microRNA and internal reference RNA for subsequent data standardization and risk assessment; B4: Data Transformation and Input; The Ct values ​​obtained by qRT-PCR detection reflect the relative expression level of each microRNA; Since the constructed prediction model is based on the GEO chip expression matrix, the Ct values ​​obtained from actual detection need to be converted into equivalent expression values ​​using Formula 1. log2(expression)=(29.67 Ct) / 0.94 (formula 1), Where 29.67 and 0.94 are the intercept and slope obtained by regression fitting of the standard curve, respectively; The transformed log2(expression) value is input into the final random forest model obtained by any one of claims 1-6 for analysis.

9. A kit for predicting the risk of AMI according to claim 8, characterized in that, The reaction system for RT-qPCR amplification is as follows: The amplification program was fixed as follows: initial denaturation at 95℃ for 5 minutes, followed by 40-45 cycles. Each cycle included denaturation at 95℃ for 15 seconds and annealing / extension at 60℃ for 60 seconds. Fluorescence signals were recorded during the cycle, and the amplification curve of the target sequence was detected in real time.

10. A kit for predicting the risk of AMI according to claim 8, characterized in that, The reverse transcription system contains reverse transcriptase, dNTPs, RNase, inhibitors, and buffer. The primer binding, reverse transcription, and enzyme inactivation are completed by setting the required temperature program to generate a specific cDNA template.

Images