Methods for diagnosing and predicting the prognosis of chronic heart failure

Abstract

To provide methods for determining whether a subject has or is at risk of having heart failure, particularly heart failure with a reduced left ventricular ejection fraction (HFREF) and a heart failure with a preserved left ventricular ejection fraction (HFPEF).SOLUTION: The methods comprise determining the level of selected miRNA(s) observed in a sample obtained from the subject, where an altered level of the miRNA(s) compared to a control indicates that the subject has heart failure or is at risk of developing heart failure. Also included are methods of determining an altered risk of death or disease progression to hospitalization and to death based on alteration of selected miRNAs in a sample from the subject, and kits thereof.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] Cross - reference to Related Applications This application claims the benefit of priority of Singapore Patent Application No. 10201503644Q, filed on May 8, 2015, the content of which is hereby incorporated by reference in its entirety for all purposes.

[0002] The present invention relates generally to the field of molecular biology. The present invention particularly relates to the use of biomarkers for the detection and diagnosis of heart failure.

Background Art

[0003] Cardiovascular diseases, including heart failure, are a major health problem and account for approximately 30% of human deaths worldwide [1]. Heart failure is also the leading cause of hospitalization for adults over 65 years of age worldwide [2]. Middle - aged adults have a 20% risk of developing heart failure during their lifetime. Despite the progress of treatment methods, the morbidity and mortality rates of heart failure (about 50% over 5 years) remain high, and approximately 2% of the healthcare budget is spent on it in many economic activities [3 - 6]. The morbidity rate of heart failure will increase due to the aging of the population and the increase in major risk factors (such as diabetes, obesity, increased initial survival rate in acute myocardial infarction, severe hypertension).

[0004] Heart failure has traditionally been considered a failure of systolic function, and left ventricular ejection fraction (LVEF) has been widely used to clarify systolic function, assess prognosis, and select patients for therapeutic intervention. However, it is recognized that heart failure can occur even in the presence of normal or near-normal ejection fraction (EF). This type of heart failure is so-called “heart failure with preserved left ventricular ejection fraction (HFPEF)” and accounts for a large proportion of clinical cases of heart failure [7-9]. Heart failure with severe diastolic and / or markedly reduced EF, so-called “heart failure with reduced left ventricular ejection fraction (HFREF)”, is the best-understood type of heart failure in terms of pathophysiology and treatment

[10] . There are some epidemiological differences between patients with HFREF and patients with HFPEF. The latter are generally older, more often female, less likely to have coronary artery disease (CAD), and more likely to have underlying hypertension [7, 8, 11]. In addition, patients with HFPEF, like patients with HFREF, do not derive any clinical benefit from angiotensin-converting enzyme inhibition or angiotensin receptor blockade [12, 13]. Symptoms of heart failure may progress and suddenly appear as "acute heart failure," requiring hospitalization, although symptoms may also progress gradually.

[0005] Timely diagnosis, classification of heart failure subtypes (HFREF or HFPEF), and improved risk stratification are crucial for the management and treatment of heart failure. Therefore, there is a need to provide methods for determining the risk of heart failure progression in subjects. There is also a need to provide methods for classifying heart failure subtypes. [Overview of the project]

[0006] In one aspect, a method is provided to determine whether a subject has heart failure or is at risk of heart failure progression. In some examples, the method includes a) measuring the level of at least one miRNA from the list of “elevated” miRNAs listed in Table 25, Table 20, Table 21, or Table 22, or the level of at least one miRNA from the list of “decreased” miRNAs, in a sample taken from the subject. In some examples, the method further includes b) determining whether the level is different from that of a control, and whether the change in the miRNA level indicates that the subject has heart failure or is at risk of heart failure progression.

[0007] In another aspect, a method is provided for determining whether a subject has heart failure. In some examples, heart failure is selected from the groups consisting of heart failure with reduced left ventricular ejection fraction (HFREF) and heart failure with maintained left ventricular ejection fraction (HFPEF). In some examples, the method includes a) measuring the level of at least one miRNA listed in Table 9 in a sample obtained from the subject. In some examples, the method further includes determining whether the level of at least one miRNA indicates that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), or that the subject is at risk of progression of heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF).

[0008] Another aspect provides a method for determining the risk of heart failure patients having an altered risk of death. In some examples, this method includes a) measuring the level of at least one miRNA listed in Table 14 in a sample taken from the subject. In some examples, this method also includes b) measuring the level of at least one miRNA listed in Table 14. In some examples, this method also includes c) determining whether the level of at least one miRNA listed in Table 14 is different from the level of miRNAs in a control population, where the change in miRNA level indicates that the subject is more likely to have an altered risk of death (altered measured overall survival) compared to the control population.

[0009] In yet another aspect, a method is provided for determining the risk of heart failure patients having an altered risk of disease progression leading to hospitalization or death. In some examples, the method includes a) measuring the level of at least one miRNA listed in Table 15 in a sample taken from the subject. In some examples, the method includes b) measuring the level of at least one miRNA listed in Table 15. In some examples, the method includes c) determining whether the level of at least one miRNA listed in Table 15 is different from the level of miRNAs in a control population, and that the change in miRNA levels indicates that the subject is more likely to have an altered risk of disease progression leading to hospitalization or death compared to the control population.

[0010] Another aspect provides a method for determining the risk of a subject developing heart failure or whether the subject already has heart failure. In some examples, the method includes (a) a step of detecting the presence of miRNAs in a sample taken from the subject. In some examples, the method further includes (b) a step of measuring the levels of at least three miRNAs listed in Table 16 or Table 23. In some examples, the method further includes (c) a step of predicting the likelihood that the subject will develop or already has heart failure, using a score based on the levels of miRNAs measured in step (a).

[0011] Another aspect provides a method for determining the risk of a subject developing heart failure or whether the subject already has heart failure. In some examples, the method includes (a) a step of detecting the presence of miRNAs in a sample taken from the subject. In some examples, the method also includes (b) a step of measuring the levels of at least three miRNAs listed in Table 17 in the sample. In some examples, the method also includes (c) a step of predicting the likelihood that the subject will develop or already has heart failure, using a score based on the levels of miRNAs measured in step (a).

[0012] Another aspect provides a method for determining whether a subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF). In some examples, the method includes (a) a step of detecting the presence of miRNAs in a sample taken from the subject. In some examples, the method includes (b) a step of measuring the levels of at least three miRNAs listed in Table 18 in the sample. In some examples, the method includes (c) a step of predicting whether the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), using a score based on the levels of miRNAs measured in step (a).

[0013] In yet another aspect, a method is provided for determining whether a subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF). In some examples, the method includes (a) the step of detecting the presence of miRNAs in a sample taken from the subject. In some examples, the method includes (b) the step of measuring the levels of at least three miRNAs listed in Table 19 or Table 24 in the sample. In some examples, the method also includes (c) the step of predicting whether the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), using a score based on the levels of miRNAs measured in step (a).

[0014] The present invention will be better understood when viewed in conjunction with non-limiting examples and the accompanying drawings in the detailed description. [Brief explanation of the drawing]

[0015] [Figure 1] Figure 1 shows a schematic diagram summarizing the number of miRNAs identified from the studies described herein.

[0016] [Figure 2]Figure 2 shows the histogram and distortion plot of the N-terminal prohormone (NT-proBNP) of brain natriuretic peptide (NATR) and the natural logarithm of the N-terminal prohormone level of NTR (ln_NT-proBNP). The distribution of NT-proBNP levels (A-C) and ln_NT-proBNP levels (natural logarithm of NT-proBNP, D-F) are shown for control patients (A, D), patients with heart failure with reduced left ventricular ejection fraction (HFREF) (B, E), and patients with heart failure with preserved left ventricular ejection fraction (HFPEF) (C, F). The skewness of each graph is calculated and displayed. Figure 2 shows that the N-terminal prohormone (NT-proBNP) of brain natriuretic peptide was positively skewed in all groups. In contrast, the natural logarithm of the N-terminal prohormone level of NTR (ln_NT-proBNP) is less skewed. Therefore, in all analyses involving NT-proBNP, we used the natural logarithm of the N-terminal prohormone level of the brain natriuretic peptide.

[0017] [Figure 3]Figure 3 shows the results of an analysis of the performance of the natural logarithm of the N-terminal prohormone of brain natriuretic peptide (ln_NT-proBNP) as a biomarker for heart failure. Individually, (A) shows a box plot representation of the ln_NT-proBNP (natural logarithm of the N-terminal prohormone of brain natriuretic peptide) levels. Each box plot represents the 25th, 50th, and 75th percentiles in the distribution. (B-D) show the receiver operating characteristic curves of ln_NT-proBNP, comparing control and heart failure (HFREF and HFPEF) (B), HFREF and cardiac HFPEF (C), control and HFREF (D), and control and HFPEF (E). AUC: Area below the receiver operating characteristic curve, C: Control (healthy), HF: Heart failure, HFREF: Subjects with reduced left ventricular ejection fraction, HFPEF: Subjects with maintained left ventricular ejection fraction. Figure 3A shows that the performance degradation of the NT-proBNP test is more pronounced in HFPEF. Figures 3B to 3D show that the natural logarithm of the N-terminal prohormone of brain natriuretic peptide (ln_NT-proBNP) was superior to HFPEF in detecting HFREF.

[0018] [Figure 4]Figure 4 shows an example workflow for high-throughput miRNA RT-PCR measurement. The steps shown in Figure 4 include isolation, multiplexing, multiplex RT, amplification, singleplex PCR, and synthetic miRNA standard curves. Details of each step are as follows: Isolation is the process of isolating and purifying miRNA from a plasma sample; spike-in miRNA refers to non-natural synthetic miRNA mimetics (small single-stranded RNAs ranging in length from 22 to 24 nucleotides) added to the sample to monitor efficiency in each step (including isolation, reverse transcription, amplification, and qPCR); multiplexing refers to a miRNA assay that is intentionally divided in silico into a large number of multiplexing groups (45 to 65 miRNAs per group) to minimize nonspecific amplification and primer-primer interactions between the RT and amplification processes; multiplexing refers to a variety of reverse transcription primers combined and added to different multiplexing groups to produce cDNA. Amplification means combining a pool of PCR primers and adding them to each cDNA pool created from a predetermined multiple group, then performing optimized touchdown PCR to simultaneously increase the total amount of cDNA in each group. Singleplex qPCR means distributing the amplified cDNA pool into various wells in a 384-well plate, and then performing a singleplex qPCR reaction. The synthetic miRNA standard curve refers to the synthetic miRNA standard curve measured with the sample to interpolate the absolute copy number throughout the entire measurement.

[0019] [Figure 5]Figure 5 shows the results of principal component analysis as a bar graph. The principal component analysis was performed based on the log2-scale expression levels (copies / ml) for 137 reliably detected mature miRNAs (Table 4). (A): Eigenvalues of the top 15 principal components. (B) Classification efficiency (AUC) of the top 15 principal components for separating control (C) and heart failure (HF). (C) Classification efficiency (AUC) of the top 15 principal components for separating HFREF (heart failure with reduced left ventricular ejection fraction) and HFPEF (heart failure with preserved left ventricular ejection fraction). AUC: Area under the receiver operating characteristic curve. Figure 5 shows that classification of HFREF and HFPEF may require capturing multi-dimensional information with a multivariate assay.

[0020] [Figure 6] Figure 6 shows a scatter plot comparing the top (AUC) principal components of heart failure subjects with controls. Viewed individually, the top (AUC) principal components used for discriminating between control (C, black circles) and heart failure (HF, white triangles) subjects are shown in A. The top (AUC) two principal components for discriminating HFREF (heart failure with reduced left ventricular ejection fraction) from HFPEF (heart failure with preserved left ventricular ejection fraction) are shown in B. AUC: Area under the receiver operating characteristic curve. PC: Number of principal components based on Figure 10. Variance: Proportion of variance represented by the principal components calculated by the eigenvalues. Figure 6 shows that controls, HFREF subjects, and HFPEF subjects can be separated based on their miRNA profiles.

[0021] [Figure 7]Figure 7 is a Venn diagram showing the overlap of biomarkers that may be used for the detection of heart failure. Comparisons between various groups (HF, HFREF, HFPEF) of controls (healthy) and heart failure patients were performed by univariate analysis (t-test) and multivariate analysis (logistic regression) incorporating age and AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. For three comparisons: C vs HF (HFREF and HFPEF), C vs HFREF, and C vs HFPEF, the number and overlap of miRNAs with p-values less than 0.01 in univariate analysis (A) and multivariate analysis (B) are shown. HF: heart failure, HFPEF: heart failure with preserved left ventricular ejection fraction, HFREF: heart failure with reduced left ventricular ejection fraction, C: control (healthy). Figure 7 shows that many miRNAs were found to be different between the control and two heart failure subtypes. Thus, it is proven that there is a true difference between these two subtypes with respect to miRNA expression.

[0022] [Figure 8] Figure 8 shows box plots and receiver operating characteristic curves of the top-ranked upregulated and downregulated miRNAs between healthy controls and heart failure patients. These box plots and receiver operating characteristic curves (ROC) are for upregulated miRNAs (A: ROC curve, C: box plot) and downregulated miRNAs (B: ROC curve, D: box plot) that ranked higher (based on AUC) in all heart failure patients compared to control (healthy) patients. The expression levels of miRNAs (copies / ml) were shown on a log2 scale. This box plot shows the 25th percentile, 50th percentile, and 75th percentile in the distribution of expression levels. C: control (healthy), HF: heart failure. AUC: area under the curve below the receiver operating characteristic curve. Figure 8 shows that combining multiple miRNAs may improve the performance of heart failure diagnosis.

[0023] [Figure 9]Figure 9 is a Venn diagram showing the overlap of biomarkers for detecting heart failure and classifying heart failure subtypes. Comparisons of HFREF and HFPEF were performed using univariate analysis (t-test) and multivariate analysis (logistic regression) incorporating age, sex, BMI (body mass index), AF (atrial fibrillation or atrial flutter), and hypertension (p-value, ln_BNP). miRNAs with p-values ​​less than 0.01 (after adjusting for false detection rate) in univariate analysis (A) and multivariate analysis (B) were compared to miRNAs for detecting heart failure (C vs. HF, or C vs. HFREF, or C vs. HFPEF, Figure 5). HF: heart failure, HFPEF: heart failure with maintained left ventricular ejection fraction, HFREF: heart failure with reduced left ventricular ejection fraction, C: control (healthy) subjects.

[0024] [Figure 10] Figure 10 shows box plots and receiver operating characteristic curves (ROCs) of the most frequently upregulated and downregulated miRNAs in HFPEF patients compared to HFREF patients. These box plots and ROCs represent the upregulated miRNAs (A: ROC curve, C: box plot) and downregulated miRNAs (B: ROC curve, D: box plot) that were more frequently upregulated (based on AUC) in HFPEF patients compared to HFREF patients. miRNA expression levels (copies / ml) are shown on a log2 scale. The box plots show the 25th, 50th, and 75th percentiles in the expression level distribution. HFPEF: heart failure with maintained left ventricular ejection fraction, HFREF: heart failure with reduced left ventricular ejection fraction, AUC: area below the receiver operating characteristic curve. Figure 10 suggests that combining multiple miRNAs in a multivariate exponential assay may provide greater diagnostic power for subtype classification.

[0025] [Figure 11]Figure 11 shows a line graph of overlapping miRNAs for the detection of heart failure and classification of heart failure subtypes. 38 overlapping miRNAs (Figure 7, A) between control, heart failure (HFREF or HFPEF), HFREF, and HFPEF were separated into seven groups based on variation. Two groups were defined as equal if the p-value (t-test) of a miRNA after a false detection test was greater than 0.01. Expression levels were based on a log2 scale and standardized to a mean of zero for each miRNA. HFPEF: heart failure with maintained left ventricular ejection fraction, HFREF: heart failure with reduced left ventricular ejection fraction, C: control (healthy). Figure 11 shows that HFPEF had a distinctly different miRNA profile compared to the HFREF subtype, unlike LVEF and NT-proBNP when compared to a healthy control. Figure 11 suggests that miRNAs may complement NT-proBNP in better identifying HFPEF.

[0026] [Figure 12] Figure 12 shows a scattering plot of correlations between all reliably detected miRNAs. Based on expression levels (copies / ml) on a log2 scale, Pearson linear correlation coefficients were calculated between all 137 reliably detected miRNAs (Table 4). Each point represents a pair of miRNAs with a correlation coefficient greater than 0.5 (A, positive correlation) or less than -0.5 (B, negative correlation). MiRNAs that differ in expression between C and HFREF and HFPEF are shown in black in the horizontal dimension. HF: heart failure, HFPEF: heart failure with maintained left ventricular ejection fraction, HFREF: heart failure with reduced left ventricular ejection fraction, C: control (healthy). Figure 12 shows that many pairs of miRNAs were similarly regulated among all subjects.

[0027] [Figure 13]Figure 13 is a bar graph representing drug therapies for HFREF and HFPEF. It summarizes the number of cases treated with various anti-HF drugs for the 327 subjects included in the prognostic analysis, and is shown separately for HFREF subtype and HFPEF subtype. A chi-square test was applied to compare the two subtypes for each treatment method. *: p value < 0.05, **: p value < 0.01, ***: p value < 0.001. Figure 13 is a summary of treatments currently being used in clinical practice and was included as a clinical variable for analyzing prognostic markers.

[0028] [Figure 14A] Figure 14 shows the survival analysis of the subjects. Individually, (A) shows Kaplan-Meier plots of clinical variables (Table 14) for which the prediction of observed survival rate based on univariate analysis (p value < 0.05) is significant. For classification variables, the positive group (black) and the negative group (gray) were compared. For normally distributed variables, subjects with values ​​above the median (black) and subjects with values ​​below the median (gray) were compared. A log-rank test was performed to test the two groups for each variable, and the p values ​​are shown above each plot. (B) is a bar graph showing the observed survival rate (OS) at 750 days after treatment. [Figure 14B] Figure 14 shows the survival analysis of the subjects. Individually, (A) shows Kaplan-Meier plots of clinical variables (Table 14) for which the prediction of observed survival rate based on univariate analysis (p value < 0.05) is significant. For classification variables, the positive group (black) and the negative group (gray) were compared. For normally distributed variables, subjects with values ​​above the median (black) and subjects with values ​​below the median (gray) were compared. A log-rank test was performed to test the two groups for each variable, and the p values ​​are shown above each plot. (B) is a bar graph showing the observed survival rate (OS) at 750 days after treatment.

[0029] [Figure 15A]Figure 15 shows the survival analysis for event-free survival. Individually, (A) shows Kaplan-Meier plots of clinical variables (Table 14) for which the prediction of event-free survival rate based on univariate analysis (p value < 0.05) is significant. For classification variables, positive groups (black) and negative groups (gray) were compared. For normally distributed variables, subjects with values ​​above the median (black) and subjects with values ​​below the median (gray) were compared. A log-rank test was performed to test the two groups for each variable, and the p values ​​are shown above each plot. (B) is a bar graph showing the event-free survival rate (EFS) at 750 days post-treatment. [Figure 15B] Figure 15 shows the survival analysis for event-free survival. Individually, (A) shows Kaplan-Meier plots of clinical variables (Table 14) for which the prediction of event-free survival rate based on univariate analysis (p value < 0.05) is significant. For classification variables, positive groups (black) and negative groups (gray) were compared. For normally distributed variables, subjects with values ​​above the median (black) and subjects with values ​​below the median (gray) were compared. A log-rank test was performed to test the two groups for each variable, and the p values ​​are shown above each plot. (B) is a bar graph showing the event-free survival rate (EFS) at 750 days post-treatment.

[0030] [Figure 16] Figure 16 shows a Venn diagram comparing biomarkers related to observed survival (OS) and event-free survival (EFS). Individually, (A) shows a comparison between miRNAs identified by univariate and multivariate analysis using the CoxPH model that are significant in OS prognosis. (B) shows a comparison between miRNAs significant in OS prognosis and miRNAs significant in EFS prognosis. miRNAs were identified by univariate or multivariate analysis using the CoxPH model. Figure 16 shows that the mechanisms differ between death and recurrent decompensated heart failure.

[0031] [Figure 17]Figure 17 shows a Venn diagram comparing biomarkers related to observed survival (OS) and event-free survival (EFS). Individually, (A) shows a comparison between miRNAs that are significantly prognostic according to the CoxPH model (for OS or EFS) and miRNAs that are significant in detecting HF (either subtype). All miRNAs were identified by univariate or multivariate analysis. (B) shows a comparison between miRNAs that were identified as significantly prognostic according to the CoxPH model (for OS or EFS) and miRNAs that are significant in classifying the two HP subtypes. All miRNAs were identified by univariate or multivariate analysis. Figure 17 shows that many of the prognostic markers were not found in the other two lists. This suggests that assays for prognosis can be formed by using a different set of miRNAs or by combining a different set of miRNAs.

[0032] [Figure 18A] Figure 18 shows the analysis of miRNAs with the highest and lowest hazard ratios for observed survival (OS). In (A), univariate or multivariate CoxPH models were constructed using the miRNA with the highest hazard ratio (hsa-miR-503) and the miRNA with the lowest hazard ratio (hsa-miR-150-5p) for observed survival (OS). These models included six additional clinical variables related to observed survival (OS): sex, hypertension, BMI, ln_NT-proBNP, β-blockers, and warfarin. The scales of the normal variables, including BMI, ln_NT-proBNP, and miRNA expression levels (log2 scale), were changed so that all levels had one standard deviation. Based on the explanatory score values ​​from the CoxPH model, the top 50% of subjects (black) and the bottom 50% of subjects (gray) were compared. A log-rank test was performed to test the two groups, and the p-values ​​were shown. (B) represents the observed overall survival (OS) at 750 days after treatment. [Figure 18B]Figure 18 shows the analysis of miRNAs with the highest and lowest hazard ratios for observed survival (OS). In (A), univariate or multivariate CoxPH models were constructed using the miRNA with the highest hazard ratio (hsa-miR-503) and the miRNA with the lowest hazard ratio (hsa-miR-150-5p) for observed survival (OS). These models included six additional clinical variables related to observed survival (OS): sex, hypertension, BMI, ln_NT-proBNP, β-blockers, and warfarin. The scales of the normal variables, including BMI, ln_NT-proBNP, and miRNA expression levels (log2 scale), were changed so that all levels had one standard deviation. Based on the explanatory score values ​​from the CoxPH model, the top 50% of subjects (black) and the bottom 50% of subjects (gray) were compared. A log-rank test was performed to test the two groups, and the p-values ​​were shown. (B) represents the observed overall survival (OS) at 750 days after treatment.

[0033] [Figure 19A] Figure 19 shows the analysis of miRNAs with the highest and lowest hazard ratios for EFS. In (A), a univariate or multivariate CoxPH model was constructed using the miRNA with the highest hazard ratio for EFS (hsa-miR-331-5p) and the miRNA with the lowest hazard ratio (hsa-miR-191-5p). This model included two additional clinical variables: diabetes status and ln_NT-proBNP related to EFS. The scales were changed so that all levels of the normal variables, including diabetes status, ln_NT-proBNP, and miRNA expression levels (log2 scale), had a value of 1 standard deviation. Based on the explanatory score values ​​from the CoxPH model, the top 50% of subjects (black) and the bottom 50% of subjects (gray) were compared. A log-rank test was performed to test the two groups, and the p-values ​​are shown. (B) shows the EFS at 750 days post-treatment. [Figure 19B]Figure 19 shows the analysis of miRNAs with the highest and lowest hazard ratios for EFS. In (A), a univariate or multivariate CoxPH model was constructed using the miRNA with the highest hazard ratio for EFS (hsa-miR-331-5p) and the miRNA with the lowest hazard ratio (hsa-miR-191-5p). This model included two additional clinical variables: diabetes status and ln_NT-proBNP related to EFS. The scales were changed so that all levels of the normal variables, including diabetes status, ln_NT-proBNP, and miRNA expression levels (log2 scale), had a value of 1 standard deviation. Based on the explanatory score values ​​from the CoxPH model, the top 50% of subjects (black) and the bottom 50% of subjects (gray) were compared. A log-rank test was performed to test the two groups, and the p-values ​​are shown. (B) shows the EFS at 750 days post-treatment.

[0034] [Figure 20A] Figure 20 shows representative results for generating a multivariate biomarker panel for detecting heart failure. In (A), the box plot shows the diagnostic power (AUC) of the multivariate biomarker panel (number of miRNAs = 3-10) in the detection and confirmation phases of heart failure when performing in silico two-fold cross-confirmation. The box plots represent the 25th, 50th, and 75th percentiles of the AUC for classifying healthy individuals and heart failure patients. Quantitative representation of the results for the detection set (black) and confirmation set (gray) is shown in (B). Error bars represent the standard deviation of the AUC. A right-handed t-test was performed to compare all adjacent gray bars to test the significance of the AUC improvement in the confirmation set when more miRNAs were included in the panel. *: p-value < 0.05, **: p-value < 0.01, ***: p-value < 0.001. [Figure 20B]Figure 20 shows representative results for generating a multivariate biomarker panel for detecting heart failure. In (A), the box plot shows the diagnostic power (AUC) of the multivariate biomarker panel (number of miRNAs = 3-10) in the detection and confirmation phases of heart failure when performing in silico two-fold cross-confirmation. The box plots represent the 25th, 50th, and 75th percentiles of the AUC for classifying healthy individuals and heart failure patients. Quantitative representation of the results for the detection set (black) and confirmation set (gray) is shown in (B). Error bars represent the standard deviation of the AUC. A right-handed t-test was performed to compare all adjacent gray bars to test the significance of the AUC improvement in the confirmation set when more miRNAs were included in the panel. *: p-value < 0.05, **: p-value < 0.01, ***: p-value < 0.001.

[0035] [Figure 21] Figure 21 shows a comparison of multivariate miRNA scores and NT-proBNP when detecting heart failure using a two-dimensional plot. (A) shows a two-dimensional plot of NT-proBNP levels (y-axis) and one of the six miRNA panel scores (x-axis) for all subjects. The NT-proBNP threshold (125) is shown by a dotted line. There was a mix of false-positive and false-negative subjects based on NT-proBNP. (B) shows a two-dimensional plot of NT-proBNP levels (y-axis) and the six miRNA panel scores (x-axis) to identify false-positive and false-negative subjects classified by NT-proBNP using a threshold of 125 pg / ml. The threshold miRNA score (0) is shown by a dotted line. Control subjects are indicated by a +. HFREF subjects are indicated by black circles, and HFPEF subjects are indicated by white triangles. Figure 21 confirms the hypothesis that miRNA biomarkers carry different information than the N-terminal prohormone of brain natriuretic peptide (NT-proBNP).

[0036] [Figure 22]Figure 22 shows the analysis of a multivariate biomarker panel for detecting heart failure by combining miRNA and NT-proBNP. (A) shows a series of box plots of the diagnostic power (AUC) of the multivariate biomarker panel (ln_NT-proBNP and 2-8 miRNAs) in the detection and confirmation phases of heart failure when two-fold cross-confirmation is performed in silico. The box plots represent the 25th, 50th, and 75th percentiles of AUC for classifying healthy individuals and HF patients. (B) shows the quantitative representation of the results for the detection set (black) and confirmation set (gray), and ln_NT-proBNP itself (first column). The error bars represent the standard deviation of the AUC. A right-handed t-test was performed to compare all adjacent gray bars to test the significance of the AUC improvement in the confirmation set when more miRNAs were included in the panel. *: p-value < 0.05, **: p-value < 0.01, ***: p-value < 0.001. Therefore, Figure 22 shows that combining miRNA with the N-terminal prohormone of brain natriuretic peptide (NT-proBNP) significantly improves classification efficiency.

[0037] [Figure 23] Figure 23 shows a Venn diagram of the overlap of miRNAs selected in a multivariate heart failure detection panel with and without the addition of the brain natriuretic peptide N-terminal prohormone (NT-proBNP). The diagram compares biomarkers selected using only miRNAs (Table 16) or miRNAs and NT-proBNP (Table 17) to detect heart failure in the multivariate biomarker discovery process. Significant miRNAs (A) and non-significant miRNAs (B) were compared separately. Figure 23 demonstrates that different miRNA lists can be used when NT-proBNP is employed.

[0038] [Figure 24]Figure 24 shows representative results for generating miRNA panels to stratify heart failure subtypes with and without NT-proBNP. (A) shows a multivariate miRNA biomarker panel search (3-10 miRNAs) for classifying heart failure subtypes. AUC results for the discovery set (black bars) and confirmation set (gray bars) are shown. (B) shows a multivariate miRNA + NT-proBNP biomarker panel search (ln_NT-proBNP and 2-8 miRNAs) for classifying heart failure subtypes. Error bars represent the standard deviation of AUC. Right-handed t-tests were performed to compare all adjacent gray bars. *: p-value < 0.05, **: p-value < 0.01, ***: p-value < 0.001. Figure 24 shows that using both miRNA and NT-proBNP may enable a clearer classification.

[0039] A brief explanation of the table The present invention will be better understood when viewed in conjunction with non-limiting examples and the attached tables in the detailed description.

[0040] Table 1 summarizes the serum / plasma miRNA biomarkers reported for heart failure. Studies measuring cell-free serum / plasma miRNA or whole blood were included in this table. Only miRNAs confirmed by qPCR are shown. Upregulated: miRNAs at higher levels in heart failure patients than in control (healthy) subjects. Downregulated: miRNAs at lower levels in heart failure patients than in control (healthy) subjects. The numbers in "Study Design" indicate the number of samples used in the study. PBMC: Peripheral blood monocytes, AMI: Acute myocardial infarction, HF: Heart failure, HFPEF: Heart failure with preserved left ventricular ejection fraction, HFREF: Heart failure with reduced left ventricular ejection fraction, BNP: Brain natriuretic peptide, C: Control (healthy subjects).

[0041] Table 2 lists the clinical information of the subjects included in the study. Clinical information of 546 subjects was included in the study. All plasma samples were stored at -80°C until use. NA: Not available, C: Control (healthy subject), PEF: Heart failure with maintained left ventricular ejection fraction, REF: Heart failure with reduced left ventricular ejection fraction.

[0042] Table 3 shows the characteristics of healthy subjects and heart failure patients. Ejection fraction (left ventricular ejection fraction), ln_NT-proBNP, age, and body mass index are shown as arithmetic mean ± standard deviation, and NT-proBNP is shown as geometric mean. The percentage value after the variable name indicates the percentage of subjects for whom the value of that variable is known. HF: heart failure, HFPEF: heart failure with maintained left ventricular ejection fraction, HFREF: heart failure with reduced left ventricular ejection fraction, C: control (healthy subjects). To compare variables between control and heart failure (C vs HF) and between HFPEF and HFREF (HFPEF vs HFREF), a t-test was used for normal variables and a chi-squared test was used for categorical variables.

[0043] Table 4 lists the sequences of 137 mature miRNAs that were reliably detected. These 137 mature miRNAs were reliably detected in plasma samples. "Reliably detected" was defined as having concentrations greater than 500 copies / ml in at least 90% of the plasma samples. The miRNAs were named according to their release by miRBase V18.

[0044] Table 5 lists the miRNAs whose expression differed between controls and all heart failure subjects. Comparisons between controls (healthy) and all heart failure subjects (both HFREF and HFPEF) were performed using univariate analysis (p-value, t-test) and multivariate analysis (p-value, logistic regression) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. The enhancement of the diagnostic performance of ln_NT-proBNP for heart failure by miRNAs was tested using logistic regression (p-value, ln-BNP) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. All p-values ​​were adjusted using the Bonferroni method to correct for false detection rates. Only miRNAs with p-values ​​less than 0.01 in both the "p-value, t-test" and "p-value, logistic regression" tests are shown. Multiplier change: The value obtained by dividing the miRNA expression level of heart failure subjects by the miRNA expression level of control subjects.

[0045] Table 6 lists the miRNAs whose expression differed between the control and HFREF. Comparison between the control (healthy) and HFREF (heart failure with reduced left ventricular ejection fraction) was performed using univariate analysis (p-value, t-test) and multivariate analysis (p-value, logistic regression) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. The enhancement of the discriminatory power of ln_NT-proBNP for HFREF by miRNA was tested using logistic regression (p-value, ln-BNP) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. All p-values ​​were adjusted using the Bonferroni method to correct for false detection rates. Only miRNAs with a p-value less than 0.01 in both the "p-value, t-test" and "p-value, logistic regression" tests are shown. Multiplier change: The value obtained by dividing the miRNA expression level of HFREF by the miRNA expression level of the control subjects.

[0046] Table 7 lists the miRNAs whose expression differed between controls and HFPF. Comparisons between controls (healthy) and HFPF (heart failure with preserved left ventricular ejection fraction) were performed using univariate analysis (p-value, t-test) and multivariate analysis (p-value, logistic regression) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. The discriminative power of ln_NT-proBNP for diagnosing HFPF was enhanced by miRNAs, which was tested using logistic regression (p-value, ln_BNP) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. All p-values ​​were adjusted using the Bonferroni method to correct for false detection rates. Only miRNAs with p-values ​​less than 0.01 in both the "p-value, t-test" and "p-value, logistic regression" tests are shown. Dilution factor: The value obtained by dividing the miRNA expression level of HFPF by the miRNA expression level of control subjects.

[0047] Table 8 shows the results of comparing this study with previously published reports. miRNAs not listed in Table 4 (expression levels less than 500 copies / ml) may not have been included in this study or were below the detection limit, and are indicated as NA (not available). Upward: miRNA expression levels are higher in heart failure patients compared to control (healthy) subjects. Downward: miRNA expression levels are lower in heart failure patients compared to control (healthy) subjects. miRNAs with a p-value of less than 0.01 after correcting for false detection rate are indicated as no change. For hsa-miR-210, the direction of change was inconsistent across various literatures (indicated as "upward and downward").

[0048] Table 9 lists miRNAs whose expression differs between HFREF and HFPEF. Comparison of HFREF (heart failure with reduced left ventricular ejection fraction) and HFPEF (heart failure with maintained left ventricular ejection fraction) was performed using univariate analysis (p-value, t-test) and multivariate analysis (p-value, logistic regression) adjusted for age, sex, BMI (body mass index), AF (atrial fibrillation or atrial flutter), and hypertension. The ability of ln_NT-proBNP to distinguish between HFREF and HFPEF was enhanced by miRNAs, which was tested using logistic regression (p-value, ln_BNP) adjusted for age, sex, BMI, AF, and hypertension. All p-values ​​were adjusted using the Bonferroni method to correct for false detection rates. Only miRNAs with p-values ​​less than 0.01 in the "p-value, t-test" tests are shown. Multiplicative change: The value obtained by dividing the miRNA expression level of HFPEF by the miRNA expression level of HFREF.

[0049] Table 10 lists the clinical information of the subjects included in the prognosis study. Clinical information of 327 subjects was included in this study. All subjects were followed for two years after being invited to the SHOP cohort study. 49 patients died during the follow-up period.

[0050] Table 11 lists the treatments used by subjects included in the prognosis study. Drug treatments were included for 327 subjects in this study. Drug names: Me1: ACE inhibitors, Me2: Angiotensin II receptor blockers, Me3: Loop / thiazide diuretics, Me4: Beta-blockers, Me5: Aspirin or Plavix, Me6: Statins, Me7: Digoxin, Me8: Warfarin, Me9: Nitrates, Me10: Calcium channel blockers, Me11: Spironolactone, Me12: Fibrates, Me13: Antidiabetic agents, Me14: Hydralazine, Me15: Iron supplements.

[0051] Table 12 shows the analysis of clinical variables in observed survival rates. Clinical parameters included in the analysis of observed survival rates using the Cox proportional hazards model included drug treatment and other variables. Age, BMI, LVEF, and ln_NT-proBNP levels were scaled to have one standard deviation. All variables were included in the multivariate analysis. Cells with p-values ​​less than 0.05 are shown in gray. ln(HR): Natural logarithm of the hazard ratio (a positive value indicated a larger variable value and a greater probability of death), SE: Standard error.

[0052] Table 13 shows the analysis of clinical variables for event-free survival. A Cox proportional hazards model was used for the clinical parameters in the analysis of event-free survival, and the levels of age, BMI, LVEF, and ln_NT-proBNP used were scaled to have one standard deviation. Drug therapy was also included. All variables were included in the multivariate analysis. Cells with p-values ​​less than 0.05 are shown in gray. ln(HR): Natural logarithm of the hazard ratio (a positive value indicated a larger variable value and a greater probability of death), SE: Standard error.

[0053] Table 14 lists the miRNAs that significantly predicted observed survival rates. Each miRNA was analyzed using a Cox proportional hazards model, employing both univariate analysis and multivariate analysis including additional clinical variables (sex, hypertension, BMI, ln_NT-proBNP, β-blockers, warfarin). All normally distributed variables, including ln_NT-proBNP, BMI, and miRNA expression levels (log2 scale), were scaled to have one standard deviation. Cells with p-values ​​less than 0.05 are shown in gray. ln(HR): natural logarithm of the hazard ratio (positive values ​​indicated a larger variable value and a greater probability of death), SE: standard error.

[0054] Table 15 lists the miRNAs that significantly predicted event-free survival. Each miRNA was analyzed using a Cox proportional hazards model, employing both univariate analysis and multivariate analysis including additional clinical variables (diabetes and ln_NT-proBNP). All normally distributed variables, including ln_NT-proBNP and miRNA expression levels (log2 scale), were scaled to have one standard deviation. Cells with p-values ​​less than 0.05 are shown in gray. ln(HR): natural logarithm of the hazard ratio (positive values ​​indicated a larger variable value and a greater probability of death), SE: standard error.

[0055] Table 16 lists the miRNAs identified in the multivariate panel search process for detecting heart failure. It lists the miRNAs selected to create biomarker panels containing 6, 7, 8, 9, and 10 miRNAs for the purpose of detecting heart failure. The prevalence was defined as the number of miRNA counts in the entire panel divided by the total number of miRNAs in the panel. To avoid miscounting of biomarkers due to inaccurate data fitting from subpopulations resulting from the randomization process in cross-confirmation analysis, panels containing the top 10% and bottom 10% of AUCs were excluded. Only miRNAs used in more than 2% of the panel are listed. Changes in miRNAs across various subtypes of heart failure were determined based on Tables 5-7.

[0056] Table 17 lists the miRNAs identified when combined with NT-proBNP in a multivariate panel search process for detecting heart failure (HF). It lists the miRNAs selected to create biomarker panels containing NT-proBNP and 3, 4, 5, 6, 7, and 8 miRNAs for the purpose of detecting heart failure. Prevalence was defined as the number of miRNA counts in the entire panel divided by the total number of miRNAs in the panel. To avoid misidentification of biomarkers due to inaccurate data fitting from subpopulations resulting from the randomization process in cross-confirmation analysis, panels containing the top 10% and bottom 10% of AUCs were excluded. Only miRNAs used in more than 2% of the panel are listed. The significance of miRNAs added to ln_NT-proBNP in identifying various subtypes of heart failure was determined based on logistic regression using the selected miRNAs and ln_NT-proBNP as predictors. The p-value for significant miRNAs after FDR correction was <0.01.

[0057] Table 18 lists the miRNAs identified in the multivariate panel search process for classifying HF subtypes. It lists the miRNAs selected to create biomarker panels containing 6, 7, 8, 9, and 10 miRNAs for the purpose of classifying heart failure (HF) subtypes. Prevalence was defined as the number of miRNA counts in the entire panel divided by the total number of panels. To avoid miscounting of biomarkers due to inaccurate data fitting from subpopulations resulting from the randomization process in cross-confirmation analysis, panels containing the top 10% and bottom 10% of AUCs were excluded. Only miRNAs used in more than 2% of the panel are listed. Changes in miRNAs between HFREF and HFPEF subtypes were determined based on Table 9.

[0058] Table 19 lists the miRNAs identified when combined with NT-proBNP in a multivariate panel search process for classifying HF subtypes. It lists the miRNAs selected to create biomarker panels containing NT-proBNP and 5, 6, 7, and 8 miRNAs for the purpose of classifying HF subtypes. Prevalence was defined as the number of miRNA counts in the entire panel divided by the total number of panels. To avoid miscounting of biomarkers due to inaccurate data fitting from subpopulations resulting from the randomization process in cross-confirmation analysis, panels containing the top 10% and bottom 10% of AUCs were excluded. Only miRNAs used in more than 2% of the panel are listed. The significance of miRNA addition to ln_NT-proBNP was determined based on logistic regression using the selected miRNAs and ln_NT-proBNP as predictors. The p-value for significant miRNAs after FDR correction was <0.01.

[0059] Table 20 lists the miRNAs identified for the detection of heart failure (HF). Comparisons between controls (healthy) and all heart failure subjects (both HFREF and HFPEF) were performed using univariate analysis (p-value, t-test) and multivariate analysis (p-value, logistic regression) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. The enhancement of the diagnostic performance of ln_NT-proBNP for heart failure by miRNAs was tested using logistic regression (p-value, ln_BNP) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. All p-values ​​were adjusted using the Bonferroni method to correct for false detection rates. Only miRNAs with p-values ​​less than 0.01 in both the "p-value, t-test" and "p-value, logistic regression" tests are shown. Multiplier change: The value obtained by dividing the miRNA expression level of HF subjects by the miRNA expression level of control subjects. Table 20 corresponds to Table 5, but differs in that the miRNAs listed in Table 20 are not part of the conventionally known miRNAs (i.e., they are not listed in Tables 1 and 8).

[0060] Table 21 lists the miRNAs identified for HFREF detection. Comparisons between controls (healthy) and HFREF (heart failure with reduced left ventricular ejection fraction) subjects were performed using univariate analysis (p-value, t-test) and multivariate analysis (p-value, logistic regression) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. The enhancement of HFREF discrimination by ln_NT-proBNP by miRNAs was tested using logistic regression (p-value, ln-BNP) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. All p-values ​​were adjusted using the Bonferroni method to correct for false detection rates. Only miRNAs with p-values ​​less than 0.01 in both the "p-value, t-test" and "p-value, logistic regression" tests are shown. Dilution factor: The value obtained by dividing the miRNA expression level of HFREF subjects by the miRNA expression level of control subjects. Table 21 corresponds to Table 6, but differs in that the miRNAs listed in Table 21 are not part of the conventionally known miRNAs (i.e., they are not listed in Tables 1 and 8).

[0061] Table 22 lists the miRNAs identified for HFPEF detection. Comparisons between control (healthy) and HFPEF (heart failure with preserved left ventricular ejection fraction) subjects were performed using univariate analysis (p-value, t-test) and multivariate analysis (p-value, logistic regression) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. The enhancement of HFPEF discrimination by ln_NT-proBNP by miRNAs was tested using logistic regression (p-value, ln-BNP) adjusted for age, AF (atrial fibrillation or atrial flutter), hypertension, and diabetes. All p-values ​​were adjusted using the Bonferroni method to correct for false detection rates. Only miRNAs with p-values ​​less than 0.01 in both the "p-value, t-test" and "p-value, logistic regression" tests are shown. Multiplier change: The value obtained by dividing the miRNA expression level of HFPEF subjects by the miRNA expression level of control subjects. Table 21 corresponds to Table 6, but differs in that the miRNAs listed in Table 21 are not part of the conventionally known miRNAs (i.e., they are not listed in Tables 1 and 8).

[0062] Table 23 lists miRNAs frequently selected for detecting heart failure in multivariate panel search processes. It includes miRNAs selected to create biomarker panels containing NT-proBNP and 6, 7, 8, 9, and 10 miRNAs for the purpose of detecting heart failure. Prevalence was defined as the number of miRNA counts in the entire panel divided by the total number of panels. To avoid misidentification of biomarkers due to inaccurate data fitting from subpopulations resulting from the randomization process in cross-confirmation analysis, panels containing the top 10% and bottom 10% of AUCs were excluded. Only miRNAs used in more than 2% of the panel are listed. Changes in miRNAs across various subtypes of heart failure (HF) were determined based on Tables 20-22. Table 23 corresponds to Table 16, but differs in that the miRNAs listed in Table 23 are not part of the conventionally known miRNAs (i.e., not those listed in Tables 1 and 8).

[0063] Table 24 lists the miRNAs frequently selected in combination with NT-proBNP to detect heart failure in the multivariate panel search process. It lists the miRNAs selected to create biomarker panels containing NT-proBNP and 3, 4, 5, 6, 7, and 8 miRNAs for the purpose of detecting HF. Prevalence was defined as the number of miRNA counts in the entire panel divided by the total number of panels. To avoid miscounting of biomarkers due to inaccurate data fitting from subpopulations resulting from the randomization process in cross-confirmation analysis, panels containing the top 10% and bottom 10% of AUCs were excluded. Only miRNAs used in more than 2% of the panel were listed. The significance of the addition of miRNAs to ln_NT-proBNP in identifying various subtypes of HF was determined based on logistic regression using the selected miRNAs and ln_NT-proBNP as predictors. The p-value for significant miRNAs after FDR correction was <0.01. Table 24 corresponds to Table 17, but differs in that the miRNAs listed in Table 24 are not part of the conventionally known miRNAs (i.e., they are not listed in Tables 1 and 8).

[0064] Table 25 lists microRNAs that may be useful, particularly for detecting heart failure. To the best of the inventor's knowledge, these miRNAs are associated solely with heart failure. The miRNAs listed in Table 25 are not part of the conventionally known miRNAs (i.e., not those listed in Tables 1 and 8).

[0065] Table 26 is a table listing examples of biomarker panels for detecting heart failure. This table provides example panel formulas, cutoff values, and performance based on given biomarkers.

[0066] Table 27 is a table listing examples of biomarker panels for detecting heart failure subtypes. This table provides example panel formulas, cutoff values, and performance based on given biomarkers. [Modes for carrying out the invention]

[0067] Timely diagnosis, accurate classification of heart failure subtypes (e.g., heart failure with reduced left ventricular ejection fraction (HFREF), heart failure with preserved left ventricular ejection fraction (HFPEF), etc.), and improved risk stratification are crucial for the management and treatment of heart failure. One attractive approach is to utilize circulating biomarkers

[14] . Established circulating biomarkers in heart failure are cardiac natriuretic peptide (BNP), its autologous counterpart, and N-terminal prohormonal brain natriuretic peptide (NT-proBNP). Both have proven useful in the diagnosis of acute heart failure and are included in any of the leading international guidelines for the diagnosis and management of heart failure because they are independently associated with prognosis at all stages of heart failure [14, 15]. However, impaired age, renal function, obesity, and atrial fibrillation impair diagnostic performance [16, 17]. In asymptomatic left ventricular failure, early symptomatic heart failure, and treated heart failure, the discriminative power of type B peptide is significantly reduced, with BNP below 100 pg / ml in half of all stable HFREF cases, and NT-proBNP lower than the value used to rule out acute symptomatic heart failure in 20% of cases

[18] . This reduction in laboratory performance is more pronounced in cases of HFPEF

[19] . Type B peptide reflects transventricular diastolic pressure and myocyte stretching (myocyte stretching is much less increased in HFPEF, where ventricular chamber volume is normal or reduced and ventricular wall is thickened, compared to HFREF, which typically has an enlarged ventricle and variability remodeling, as well as ventricular diameter, and depends on intraventricular pressure and wall thickness)

[20] . Therefore, when screening for early-stage or partially treated heart failure, and when monitoring the state of heart failure in the chronic phase, there is a need for biomarkers that supplement or replace type B peptides, but this need is not being met. This is especially true for HFPEF, where B peptide levels are lower than and often normal than HFREF

[21] . Currently, the classification of heart failure subtypes relies on imaging and the interpretation of those images by cardiologists. No biomarker-based tests are available for this purpose.Therefore, the least invasive methods possible are desirable for improving the diagnosis of heart failure and the classification of HF subtypes.

[0068] MicroRNAs (miRNAs) are small, non-coding RNAs that play a central role in regulating abnormal gene expression associated with the pathogenesis of various diseases [22–26]. Since their discovery in 1993

[27] , miRNAs are estimated to regulate more than 60% of all human genes

[28] , and many miRNAs have been identified as key players in important cellular functions such as amplification

[29] and apoptosis

[30] . The discovery of miRNAs in human serum and plasma has opened up the possibility of using circulating miRNAs as biomarkers for the diagnosis, prognosis prediction, and treatment decision-making of many diseases [31–35]. An integrated, multifaceted approach using miRNAs or a combination of miRNAs and BNP / NT-proBNPs to diagnose HF may improve diagnostics. Combinations of genomic markers (e.g., miRNAs) and protein markers (e.g., BNP / NT-proBNPs) may enhance the diagnostic power of HF compared to using BNP / NT-proBNPs alone. Recently, various attempts have been made to identify circulating cell-free miRNA biomarkers in serum or plasma to distinguish HF patients from healthy subjects [36-47] (Table 1).

[0069] [Table 1]

[0070] These studies have reported a group of miRNAs with differing regulatory states in heart failure subjects. However, there is a lack of consistency among the published work. Of the 67 miRNAs reported, only three were found to be upregulated in two or more reports. In particular, hsa-miR-210 was reported to be upregulated in HF in one report and downregulated in another (Table 1). The lack of agreement between studies may be due to a number of reasons, including the use of small sample sizes, diversity in sample selection (e.g., disease stage), and the control used, but the control used is the most important [32, 48]. Preanalytic processes, including experimental design and workflow, are crucial for biomarker identification and confirmation. Most studies to date have used high-throughput array platforms to screen a limited number of samples. This approach lacks sensitivity and reproducibility. A small target population (fewer than 10 miRNAs) is generated, identified, and further confirmed. Most studies are awaiting confirmation in larger patient populations. Another widely adopted approach is based on screening reported candidate miRNAs using quantitative real-time polymerase chain reaction (qPCR). Evaluations of different technologies, array-based and qPCR platform-based, have revealed substantial differences in the performance of these platforms for miRNA measurement, which may explain the inconsistencies observed between studies

[49] . To date, there is no consensus on specific circulating serum / plasma miRNAs that could be used as heart failure biomarkers. No previously reported miRNA profiles have been useful for classifying heart failure subtypes. Therefore, there is a need to pre-specify and build robust technological platforms for discovering and confirming heart failure biomarkers, and to ensure the reproducibility of results.

[0071] In this disclosure, a panel of circulating miRNAs was identified as a potential biomarker for heart failure. These multivariate index assays are defined by Food and Drug Administration (FDA) guidelines, which are quoted below: "Combining values ​​of multiple variables using an interpretation function to produce a single patient-specific outcome (e.g., “classification,” “score,” “index,” etc.). The results are intended to be used for the diagnosis of disease or other symptoms, or for the cure, mitigation, treatment, or prevention of disease, but because they provide results whose derivation is opaque, they cannot be independently derived or verified by the end user." Therefore, quantitative data based on highly reliable qPCR adhering to MIQE (Minimum Information for Publication of Quantitative Real-Time PCR Experiments) is a prerequisite, and the use of currently available mathematical and biostatistical tools is crucial for simultaneously determining the interrelationships of these multiple variables.

[0072] A variety of miRNA assay methods exist, including hybridization-based methods (microarrays, Northern blotting, bioluminescence), sequencing-based methods, and qPCR-based methods

[50] . Because miRNAs are small in size (approximately 22 nucleotides), the most robust technology that provides precise, reproducible, and accurate quantitative results with the greatest dynamic range is the qPCR-based platform

[51] . Currently, it is the golden standard commonly used to validate results from other techniques (e.g., sequencing and microarray data). One variation of this method is digital PCR

[52] , a newer technology based on similar principles, but which has not yet gained widespread acceptance or use.

[0073] In this study, we profiled 203 miRNAs in the plasma of 338 chronic heart failure patients (180 with HFREF and 158 with HFPEF) and 208 non-heart failure subjects (control group) using qPCR. This is the largest cohort of miRNAs reported in the literature to date for screening in heart failure. Figure 1 summarizes the number of miRNAs identified for the various approaches proposed in this study.

[0074] The inventors of this disclosure have established a well-designed workflow with multi-layered technical and sample controls. This ensures assay reliability and minimizes the possibility of crossover between contaminants and technical noise. To discover biomarkers for the diagnosis of heart failure, the inventors screened 203 miRNAs and detected 137 miRNAs expressed in all plasma samples. Of these, 75 miRNAs were identified as significantly altered between heart failure (HFREF and / or HFPEF) and control. 52 miRNAs were found to be able to distinguish HFREF from control, and 68 were found to be significantly expressed between HFPEF and control. Thus, the inventors have identified a group of miRNAs that can distinguish HFREF from HFPEF. The inventors have also identified a group of miRNAs that are dysregulated in heart failure compared to control.

[0075] In one aspect, this provides a method for determining whether a subject has heart failure or is at risk of heart failure progression. In some examples, this method includes a) measuring the level of at least one miRNA from the list of “elevated” (above control) miRNAs listed in Table 25, Table 20, Table 21, or Table 22, or the level of at least one miRNA from the list of “decreased” (below control) miRNAs, in a sample taken from the subject. In some examples, this method further includes b) determining whether the level is different from that of a control, and whether the change in the miRNA level indicates that the subject has heart failure or is at risk of heart failure progression.

[0076] [Table 2]

[0077] [Table 3]

[0078] [Table 4]

[0079] [Table 5]

[0080] [Table 6]

[0081] [Table 7]

[0082] [Table 8]

[0083] Table 9

[0084] Throughout this disclosure, the term “miRNA” refers to microRNAs, or small non-coding RNA molecules, which are found in plants, animals, and some viruses. miRNAs are known to have functions in RNA silencing and post-transcriptional regulation of gene expression. These highly conserved RNAs regulate gene expression by binding to the 3'-untranslated region (3'-UTR) of specific mRNAs. For example, each miRNA is thought to regulate a large number of genes, as hundreds of miRNA genes are predicted to exist in higher eukaryotes. A miRNA is a covalently linked group of at least 10 and no more than 35 nucleotides. In some examples, miRNAs can be 10–33 or 15–30 nucleotides long, or 17–27 or 18–26 nucleotides long. In some cases, a miRNA can consist of 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, or 35 nucleotides, without optional labeling and / or extended sequences (e.g., biotin stretches). miRNAs regulate gene expression and are encoded by genes, with the miRNA being transcribed from the DNA of those genes, but the miRNA is never translated into a protein (i.e., miRNA is non-coding RNA). Throughout this disclosure, the measured miRNAs may have at least 90%, 95%, 97.5%, 98%, or 99% sequence matching with the miRNAs listed in any of the tables presented in this disclosure.Therefore, in some cases, the measured miRNAs are at least 90%, 95%, 97.5%, 98%, or 99% sequence-matched with the miRNAs listed in any of Tables 9, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, or 25. In this specification, the term “sequence-matching” means the relationship between two or more polypeptide sequences or two or more polynucleotide sequences, i.e., a reference sequence, and a given sequence given for comparison with that reference sequence. Sequence-matching is obtained by comparing a given sequence with a reference sequence after maximizing the sequence similarity determined by the inter-strand matching of such sequences by optimally aligning the sequences. When such alignment is performed, sequence matching is checked position by position. For example, sequences “match” at a particular position if the nucleotide or amino acid residue matches at that position. The percentage sequence-matching is then obtained by dividing the total number of such positional matches by the total number of nucleotides or residues in the reference sequence. Sequence matching can be easily calculated using methods known to those skilled in the art.

[0085] Throughout this disclosure, the terms “heart failure” or “HF” refer to a complex clinical syndrome in which the heart’s pumping function is insufficient (ventricular dysfunction) and it is unable to meet the needs of the body’s vital systems and tissues. The severity of heart failure can range from mild to severe, where the subject has no limitations on physical activity, to increasingly severe, where the subject is unable to perform physical activities without discomfort. Heart failure is a progressive and chronic disease that worsens over time. In extreme cases, heart failure may necessitate a heart transplant. In some cases, subjects are judged to be at risk of progression of heart failure if they are likely to develop it in the future, which may include worsening to recurrent acute decompensated heart failure, or death in subjects who already have chronic heart failure.

[0086] Throughout this disclosure, the terms “subject” or “patient” are interchangeable to mean an individual or mammal suspected of having heart failure. A patient may be predicted (or judged, or diagnosed) to have heart failure, i.e., the disease, or not have heart failure, i.e., be healthy. A subject may also be judged to have a specific form of heart failure. In some cases, a heart failure patient may be a subject with a primary diagnosis of heart failure and / or a subject who, after 3–5 days of treatment, has improved symptoms, resolved the clinical physical signs of heart failure, and is deemed suitable for discharge. For example, a subject may also be judged to have progressive heart failure or a specific form of heart failure. It should be noted that a subject judged to be healthy, i.e., not having heart failure or a specific form of heart failure, may have another unexamined / known disease. As used herein, a subject may be any mammal (including humans and other mammals), e.g., dogs, cats, rabbits, mice, rats, monkeys. In some cases, a human may be a subject. Therefore, the miRNA from a subject can be human miRNA or miRNA from other mammals, such as animals like mice, monkeys, or rats, or it can be human miRNA or miRNA from other mammals, such as animals like mice, monkeys, or rats, as a single collection of miRNAs. As shown in Table 2 of the Experiment section, subjects of this disclosure can be of Asian descent or of Asian ethnicity. In some examples, subjects may include, but are not limited to, Asian ethnicities such as Chinese, Indians, or Malays.

[0087] In contrast, the term “control” or “control subject,” when used in the context of this invention, can mean subjects known to have heart failure, i.e., the disease (positive controls, e.g., good prognosis, poor prognosis) (samples obtained from such subjects), and / or subjects with heart failure subtype HFPEF, and / or subjects with heart failure subtype HFREF, and / or subjects known to not have heart failure, i.e., healthy (negative controls). This term can also mean subjects (samples obtained from such subjects) known to have another disease / condition. It should be noted that control subjects known to be healthy, i.e., not having heart failure, may have another disease that has not been investigated / is known to have. Therefore, in some cases, subjects without heart failure (sometimes called normal subjects) can be controls. Any mammal (including humans and other mammals), e.g., dogs, cats, rabbits, mice, rats, monkeys, etc., can be control subjects. In some cases, the control is human. In some cases, individual subjects or cohorts of subjects (samples obtained from such subjects) can be controls.

[0088] Those skilled in the art will understand that the methods described herein should not be used as a substitute for the role of a physician in diagnosing a subject's condition. A clinical diagnosis of heart failure in a subject would require a physician to analyze other available symptoms and / or other information. The methods described herein are intended to provide supporting or additional information for a physician to make a final diagnosis of a patient / subject.

[0089] Throughout this disclosure, the term “sample” means body fluid or extracellular fluid. In some examples, body fluid includes, but is not limited to, the cellular and noncellular components of amniotic fluid, milk, bronchial lavage fluid, cerebrospinal fluid, colostrum, interstitial fluid, peritoneal fluid, pleural fluid, saliva, semen, urine, tears, and whole blood (which includes plasma, red blood cells, white blood cells, serum, etc.). In some examples, body fluid may be blood, serum plasma, or plasma.

[0090] In some cases, elevated levels of miRNAs, listed as “elevated” in Table 20 or Table 25, compared to controls, indicate that the subject has heart failure or is at risk of heart failure progression.

[0091] In some cases, a decrease in miRNA levels, listed as “reduced” in Table 20 or Table 25, compared to a control, indicates that the subject has heart failure or is at risk of heart failure progression.

[0092] As used herein, the terms “miRNA level” or “miRNA level” refer to the measurement of miRNA expression levels (or miRNA expression profiles) or indicators that correlate with miRNA expression levels in a sample, when used in the context of this disclosure. miRNA expression levels can be determined by appropriate means known in the art that enable the analysis of miRNA expression levels and the comparison of samples from subjects (e.g., potentially diseased) with control subjects (e.g., reference samples). Such means include, but are not limited to, nucleic acid hybridization (e.g., microarrays), nucleic acid amplification (PCR, RT-PCR, qRT-PCR, high-throughput RT-PCR), ELISA for quantification, next-generation sequencing (e.g., ABI SOLID, Illumina genome analyzers, Roche / 454 GS FLX), and flow cytometry (e.g., LUMINEX). The sample materials measured by the above methods can include raw samples or total RNA, treated samples or total RNA, labeled total RNA, amplified total RNA, cDNA, labeled cDNA, amplified cDNA, miRNA, labeled miRNA, amplified miRNA, and any derivatives that may arise from the above RNA / DNA species. By determining the miRNA expression level, each miRNA is represented numerically. A higher value for an individual miRNA indicates a higher (expression) level of that miRNA, and a lower value indicates a lower (expression) level of that miRNA. When a value greater than the control is detected for an individual miRNA, its miRNA expression is called "elevated" or "upregulated." Conversely, when a value less than the control is detected for an individual miRNA, its miRNA expression is called "decreased" or "downregulated."

[0093] "miRNA (expression) level" as used herein refers to the expression level / expression profile / expression data of a single miRNA, or at least two, or at least three, or at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least twelve, or at least thirteen, or at least fourteen, or at least fifteen, or at least sixteen, or at least seventeen, or at least eighteen, or at least nineteen, or at least twenty, or at least twenty-one, or at least twenty-two, or at least twenty-three, or at least twenty-five, or at least twenty-six, or at least twenty-seven, or at least twenty-eight, or at least twenty-nine, or at least thirty, or at least thirty-one, or at least thirty-two, or at least thirty-three, or at least thirty-five

[0094] In some cases, a method for determining whether a subject has heart failure or is at risk of developing heart failure may include measuring changes in the levels of at least two, or at least three, or at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least twenty, or at least ten to fifty, or at least forty to sixty, or all of the miRNAs listed in Table 20. In some cases, a method for determining whether a subject has heart failure or is at risk of developing heart failure may include measuring changes in the levels of at least two, or at least three, or at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or all of the miRNAs listed in Table 25.

[0095] In some cases, elevated miRNA levels listed as “increased” compared to controls in Table 21 using the methods described herein may indicate that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or is at risk of progression to HFREF. In some cases, decreased miRNA levels listed as “decreased” compared to controls in Table 21 using the methods described herein may indicate that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or is at risk of progression to HFREF. In some cases, methods for determining whether a subject has HFREF or is at risk of developing HFREF may include measuring changes in the levels of at least two, or at least three, or at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least two to twenty, or at least ten to forty-three, or at least fifteen to forty-three, or at least thirty to forty-three, or at least forty, or all of the miRNAs listed in Table 21.

[0096] In this specification, the terms “heart failure with reduced left ventricular ejection fraction (HFREF)” and “heart failure with maintained left ventricular ejection fraction (HFPEF)” mean the same terms as commonly used in this art. For example, the term HFREF can also mean systolic heart failure. In HFREF, the myocardium does not contract effectively, and less oxygen-rich blood is pumped into the body. Conversely, the term “heart failure with maintained left ventricular ejection fraction (HFPEF)” means diastolic heart failure. In HFPEF, the myocardium contracts normally, but it does not relax when it should during ventricular filling, i.e., when the ventricles should relax.

[0097] In some cases, elevated miRNA levels listed as “increased” compared to controls in Table 22 using the methods described herein may indicate that the subject has heart failure with preserved left ventricular fraction (HFPEF) or is at risk of progression to HFPEF. In some cases, decreased miRNA levels listed as “decreased” compared to controls in Table 22 using the methods described herein may indicate that the subject has heart failure with preserved left ventricular fraction (HFPEF) or is at risk of progression to HFPEF. In some cases, methods for determining whether a subject has HFPEF or is at risk of developing HFPEF may include measuring changes in the levels of at least two, or at least three, or at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least two to twenty, or at least ten to fifty, or at least twenty to fifty, or at least twenty to fifty, or at least thirty to sixty, or at least thirty to sixty, or at least 40 to sixty, or at least 40 to sixty, or at least fourty to sixty, or all of the miRNAs listed in Table 22.

[0098] In another aspect, a method is provided to determine whether a subject has heart failure selected from the groups consisting of heart failure with reduced left ventricular ejection fraction (HFREF) and heart failure with maintained left ventricular ejection fraction (HFPEF), the method comprising the steps of a) detecting (or measuring) the level of at least one miRNA listed in Table 9 in a sample taken from the subject, and b) determining whether the level is different from that of a control, the change in the miRNA level indicating that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), or is at risk of progression of heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF).

[0099] [Table 10]

[0100] In some cases, elevated miRNA levels listed as “elevated” in Table 9 compared to controls, using the methods described herein, may indicate that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF). In some cases, decreased miRNA levels listed as “decreased” in Table 9 compared to controls, using the methods described herein, may indicate that the subject is at risk of progression of heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF).

[0101] In some examples of methods for determining whether a subject has heart failure selected from the groups consisting of heart failure with reduced left ventricular ejection fraction (HFREF) and heart failure with preserved left ventricular ejection fraction (HFPEF), the method may include measuring changes in the levels of at least two, or at least three, or at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least two to twenty, or at least ten to thirty-nine, or all of the miRNAs listed in Table 22.

[0102] In some examples of methods for determining whether a subject has heart failure, selected from groups consisting of heart failure with reduced left ventricular ejection fraction (HFREF) and heart failure with maintained left ventricular ejection fraction (HFPEF), the control can consist of subjects with either HFREF or HFP. In some examples, the control can consist of patients with HFREF, and the difference in miRNA expression listed in Table 9 indicates that the subject has HFP. In some examples, when the control is a patient with HFP, the difference in miRNA expression listed in Table 9 indicates that the subject has HFREF.

[0103] In addition, the inventors of this disclosure have also investigated the use of these miRNAs as prognostic markers. That is, the methods of this disclosure can be used to predict the risk of future events such as death or hospitalization, or the prognosis of disease progression revealed by the diagnosis. The prognosis of a patient with heart failure means predicting the likelihood of measured survival (survival without death) or event-free survival (survival without hospitalization or death). In this specification, the terms “measured survival rate,” “all-cause survival rate,” “all-cause mortality rate,” or “all causes of death” mean the measured survival rate of a subject when all causes of death are considered. This term is in contrast to “event-free survival rate (EFS),” which means the absence of hospitalization due to recurrence of heart failure (i.e., the length of time after treatment for heart failure that recurrence hospitalization due to decompensated heart failure is avoided) and the absence of any cause of death.

[0104] The inventors of this disclosure have found that in patients with chronic heart failure, there are numerous miRNAs that have been found to be excellent predictors of observed (all-cause) survival (OS) (i.e., observed survival rate due to all causes of death), or event-free survival (EFS) that is neither due to relapse hospitalization due to heart failure (i.e., the length of time after treatment for heart failure that relapse hospitalization due to decompensated heart failure is avoided) nor to all-cause death. Thus, this disclosure can also be used in a method for predicting the prognosis of a subject. Thus, in another aspect of this disclosure, a method is provided for determining the risk of a heart failure patient whose risk of death has changed (or whose observed (all-cause) survival rate has decreased). In some examples, the method includes a) measuring the level of at least one miRNA listed in Table 14 in a sample taken from the subject; and b) determining whether the level of at least one miRNA listed in Table 14 is different from the level of miRNAs in a control population, where the change in miRNA level indicates that the subject is more likely to have an altered risk of death compared to the control population.

[0105] [Table 11]

[0106] In this specification, the term “hazard ratio” means the rate or estimate of the probability per minute of “death” or “hospitalization” at a particular moment, assuming that a subject “survived” to the moment of “death” or “hospitalization,” as is commonly known in the art. This term is used to measure the magnitude of the difference between two survival curves. A hazard ratio (HR) > 1 indicates a greater risk of shorter survival times, and a hazard ratio (HR) < 1 indicates a greater risk of longer survival times. As is known in the art, hazard ratios can be calculated using the Cox proportional hazards (CoxPH) model.

[0107] In some cases, an increase in miRNA levels listed as “Hazard Ratio (HR) > 1” in Table 14 compared to the control, using the methods described herein, may indicate an increased risk of death in the subject (a reduced observed (all-cause) survival rate). In some cases, a decrease in miRNA levels listed as “Hazard Ratio (HR) > 1” in Table 14 compared to the control, using the methods described herein, may indicate a reduced risk of death in the subject (a reduced observed (all-cause) survival rate).

[0108] In some cases, an increase in miRNA levels listed as “Hazard Ratio (HR) < 1” in Table 14 compared to the control, using the methods described herein, may indicate a reduced risk of death in the subjects (increased observed (all-cause) survival rate). In some cases, a decrease in miRNA levels listed as “Hazard Ratio (HR) < 1” in Table 14 compared to the control, using the methods described herein, may indicate an increased risk of death in the subjects (increased observed (all-cause) survival rate).

[0109] In another aspect, a method is provided to determine the risk of heart failure patients having an altered risk (reduced event-free survival) of disease progression leading to hospitalization or death. In some examples, the method includes a) measuring the level of at least one miRNA listed in Table 15 in a sample taken from the subject; and b) determining whether the level of at least one miRNA listed in Table 15 is different from the level of miRNAs in a control population, where the change in miRNA level indicates that the subject is more likely to have an altered risk of disease progression leading to hospitalization or death compared to the control population.

[0110] [Table 12]

[0111] In some cases, an increase in miRNA levels listed as “Hazard Ratio (HR) > 1” in Table 15 compared to the control, using the methods described herein, may indicate an increased risk of disease progression leading to hospitalization or death (a reduced observed (all-cause) survival rate). In some cases, a decrease in miRNA levels listed as “Hazard Ratio (HR) > 1” in Table 15 compared to the control, using the methods described herein, may indicate a reduced risk of disease progression leading to hospitalization or death (a reduced observed (all-cause) survival rate).

[0112] In some cases, an increase in miRNA levels listed as “Hazard Ratio (HR) < 1” in Table 15 compared to the control, using the methods described herein, may indicate a reduced risk of disease progression leading to hospitalization or death (increased event-free survival).

[0113] In some cases, the control population or cohort of heart failure subjects may be used as a control in the methods described herein. In some cases, the control population may be a population or cohort of heart failure subjects from which the microRNA expression levels and the risk of death or disease progression in the population can be revealed. In some cases, the microRNA expression level in the control population may be the mean or median expression level of all subjects in the population (including the patient in question). In some cases, if 10% of patients in the control population die within 5 years, the risk of death within 5 years in that control population is 10%. In some cases, the control population includes heart failure patients whose risk of death or disease progression is determined using microRNA expression levels.

[0114] In some cases, in the methods described herein, a heart failure patient may be a subject with a primary diagnosis of heart failure and / or a subject who has received treatment for 3 to 5 days when symptoms have improved and is deemed suitable for discharge when the physical signs of clinical heart failure have resolved. In some cases, a patient may be a stable compensated heart failure patient who has not yet progressed to a state requiring readmission or death due to recurrent acute decompensated heart failure.

[0115] In yet another aspect, a method is provided for determining the risk of a subject developing heart failure or whether a subject has heart failure, comprising the steps of (a) measuring the levels of at least three miRNAs listed in Table 16 or Table 23 in a sample taken from the subject; and (b) predicting the likelihood that the subject will develop or have heart failure using a score based on the miRNA levels measured in step (a). In some examples, the method further comprises measuring the level of at least one miRNA listed as “non-significant” in Table 16 or Table 23, the at least one miRNA being hsa-miR-10b-5p.

[0116] Throughout this disclosure and in reference to all methods described herein, the term “score” means an integer or number that can be mathematically determined using, for example, computational models known in the art (including, but not limited to, SMV), and calculated using one of the many formulas and / or algorithms known in the art for statistical classification. Such scores are used to select one outcome from a range of possible outcomes. The appropriateness and statistical significance of such scores depend on the size and quality of the underlying dataset used to establish the outcome spectrum. For example, if a blind sample is input into an algorithm, the algorithm can calculate a score based on the information provided from the analysis of the blind sample. This results in a score for that blind sample. Based on this score, it can be determined, for example, how likely it is that the patient from whom the blind sample was obtained has heart failure or not. The ends of the spectrum can be determined logically based on the data provided, or arbitrarily at the request of the experimenter. In either case, the spectrum must be determined before examining the blind sample. As a result, a score produced by such a blind sample, for example, a number like "45," may indicate that the corresponding patient has heart failure, if the range is defined as a scale of 1 to 50, with "1" representing no heart failure and "50" representing heart failure. Therefore, the term "score" refers to a mathematical score calculated using one of the many formulas and / or algorithms known in this field for statistical classification.Examples of such formulas and / or algorithms include (statistical) classification algorithms, the selection of which algorithms can be, but is not limited to, support vector machine algorithms, logistic regression algorithms, multinomial logistic regression algorithms, Fisher's linear discriminant algorithm, quadratic classifier algorithms, perceptron algorithms, k-nearest neighbor algorithms, artificial neural network algorithms, random forest algorithms, decision tree algorithms, naive Bayesian algorithms, adaptive Bayesian network algorithms, and multiple learning algorithms combining ensemble learning methods. In another example, the classification algorithm is pre-trained using control expression levels. In some examples, the classification algorithm compares the expression levels of a subject to the expression levels of a control and returns a mathematical score that identifies which control group the subject is most likely to belong to. In some examples, the classification algorithm can compare the expression levels of a subject to the expression levels of a control and return a mathematical score that identifies which control group the subject is most likely to belong to. Examples of algorithms that can be used in this disclosure are shown below.

[0117] [Table 13]

[0118] [Table 14]

[0119] In some examples, the methods disclosed herein measure changes in the levels of at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least two to at least twenty, or at least ten to at least 45, or at least 40 to at least 50, or all of the miRNAs listed in Table 16. In some examples, the methods disclosed herein measure at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least two to at least twenty, or at least ten to at least 41, or all of the miRNAs listed in Table 23.

[0120] In some cases, the methods disclosed herein measure the level of at least one (or more) miRNAs (in a subject's plasma sample). Combining the measurements of at least one miRNA can generate a score for predicting heart failure or classifying HFREF and HFPEF subtypes. In some cases, Equation 1 can be used as the formula for generating that score. The formula is as follows:

number

[0121] Equation 1 showed that a linear model was used to predict heart failure, or to predict the classification of HFREF subtypes and HFPEF subtypes. The predictive score (only one per subject) is a number used to determine the prediction or diagnosis.

[0122] In the experimental section of this disclosure, the usefulness of the identified miRNAs for diagnosis was further statistically evaluated. Next, multivariate miRNA biomarker panels (HF panel, HFREF panel, HFPEF panel) were constructed by repeated in silico cross-validation using forward-looking floating search (SFFS)

[53] and support vector machines (SVM)

[54] . The inventors of this disclosure found that, in receiver operating characteristic (ROC) plots, several miRNAs included in the biomarker panels consistently produced AUC (Area Under Curve) values ​​of 0.92 or higher for HF detection (Figure 20, B) and AUC of 0.75 or higher for subtype classification (Figure 24, A). When used in combination with NT-proBNP, the miRNA panels showed significantly improved discriminative power and better classification accuracy for both identification and classification purposes (Figures 22, B and 24, B).

[0123] Therefore, in another aspect, a method is provided for determining the risk of a subject developing heart failure or whether a subject has heart failure, which includes (a) measuring the levels of at least two miRNAs listed in Table 17 in a sample taken from the subject; and (b) predicting the likelihood that the subject will develop or have heart failure using a score based on the levels of miRNAs measured in step (a).

[0124] [Table 15] [Table 16]

[0125] In some cases, the methods disclosed herein may further include a step of measuring the levels of brain natriuretic peptide (BNP) and / or the N-terminal prohormone of brain natriuretic peptide (NT-proBNP). In some cases, both NT-proBNP and BNP are excellent markers for predicting and diagnosing heart failure (e.g., chronic heart failure).

[0126] In some examples, the methods disclosed herein can measure altered levels of at least three, or at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least two to at least 20, or at least ten to at least 45, or at least 40 to at least 48, or all of the miRNAs listed in Table 17.

[0127] In some examples, Equation 2 can be used instead in methods disclosed herein in which BNP and / or NT-proBNP are used together with miRNA. In Equation 2, the level of NT-proBNP in the plasma sample is included in the linear model. In one example, Equation 2 is as follows:

[0128]

number

[0129] In addition, to predict heart failure, a predictive score (which is thought to be one per subject) is a number indicating the likelihood that a subject has heart failure. In some cases, the results of the methods disclosed herein (i.e., the likelihood or prediction of diagnosis) can be found in Equation 3. If the value is greater than a predetermined cutoff value, the subject is diagnosed or predicted to have heart failure. If the value is less than a predetermined cutoff value, the subject is diagnosed or predicted not to have heart failure. Equation 3 is as follows:

[0130]

number

[0131] In some cases, to classify between HFREF and HFPEF subtypes, the predictive score (which is thought to be one per subject) is a number that indicates how likely a subject is that a heart failure patient has HFPEF subtype heart failure. In some cases, the diagnostic outcome can be found in Equation 4. If the value is greater than a predetermined cutoff value, the heart failure subject is diagnosed (or predicted) to have HFPEF subtype heart failure. If the value is less than a predetermined cutoff value, the subject is diagnosed (or predicted) to have HFREF subtype heart failure.

number

[0132] In another aspect, a method is provided for determining whether a subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF). In some examples, the method includes (a) measuring the levels of at least three miRNAs listed in Table 18 in a sample taken from the subject; and (b) using a score based on the miRNA levels measured in step (a) to predict whether the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF).

[0133] [Table 17]

[0134] In some examples, the methods disclosed herein can measure altered levels of at least four, or at least five, or at least six, or at least seven, or at least eight, or at least nine, or at least ten, or at least eleven, or at least two to at least twenty, or at least ten to at least thirty, or at least forty to at least 45, or all of the miRNAs listed in Table 18.

[0135] In some cases, the score in the methods disclosed herein can be calculated by formulas provided herein. In some cases, at least one of these formulas is possible, including but not limited to formula 1 and / or formula 2. The results of the methods disclosed herein can be obtained by these formulas (e.g., formula 3, formula 4, etc., but not limited to these).

[0136] In yet another aspect, a method is provided for determining whether a subject is likely to have heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), comprising: (a) measuring the levels of at least two miRNAs listed in Table 19 or Table 24 in a sample taken from the subject; and (b) predicting whether the subject is likely to have heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF) using a score based on the miRNA levels measured in step (a).

[0137] [Table 18]

[0138] [Table 19]

[0139] As shown in the experimental section and drawings (e.g., Figures 22 and 24), when the method of this disclosure is used in conjunction with an additional step of determining NT-proBNP, the method provides remarkably accurate predictions. Therefore, in some examples, the method further includes a step of measuring the levels of brain natriuretic peptide (BNP) and / or the N-terminal prohormone of brain natriuretic peptide (NT-proBNP).

[0140] In some examples, the methods disclosed herein can measure altered levels of at least 3, or at least 4, or at least 5, or at least 6, or at least 7, or at least 8, or at least 9, or at least 10, or at least 11, or at least 2 to at least 20, or at least 10 to at least 30, or all of the miRNAs listed in Table 19, or at least 4, or at least 5, or at least 6, or at least 7, or at least 8, or at least 9, or at least 10, or at least 11, or at least 2 to at least 20, or at least 10 to at least 41, or all of the miRNAs listed in Table 24.

[0141] In some cases, the level of at least one miRNA measured in step (b) did not change in the subjects compared to the control. In such cases, the miRNAs whose levels did not change in the subjects compared to the control are listed as “not significant” in their respective tables.

[0142] In some examples, the score in the method disclosed herein can be calculated by Equation 2.

[0143] As those skilled in the art will see, the classification algorithms used in any of the methods disclosed herein can be pre-trained using the expression levels of controls. When the classification algorithm is pre-trained using existing clinical data, the controls may consist of at least one person selected from a group comprising controls without heart failure (normal) and patients with heart failure. The controls may include a cohort of subjects with and / or without heart failure (i.e., without heart failure). Thus, in some examples of the methods disclosed herein, the controls may include, but are not limited to, controls without heart failure, patients with heart failure, patients with heart failure of the HFPEF subtype, patients with heart failure of the HFREF subtype, and so on.

[0144] This disclosure discusses the comparison of differences in miRNA expression levels when establishing a miRNA panel. Based on this, it is possible to determine whether a subject is at risk of progression of heart failure or has heart failure. As disclosed herein, the methods disclosed herein typically require the comparison of differences in miRNA expression levels from different groups. For example, the comparison is made between two groups. These comparison groups can be defined as heart failure and no heart failure (normal), but are not limited to these. Further subgroups (e.g., HFREF and HFPEF) can be found within the heart failure group, but are not limited to these subgroups. The comparison of differences can also be performed between these groups as described herein. Thus, in some examples, miRNA expression levels can be expressed as concentration, logarithm (concentration), threshold cycle / quantification cycle (Ct / Cq) count, power of 2 threshold cycle / quantification cycle (Ct / Cq) count, etc., but are not limited to these.

[0145] Any method described herein may further include, but is not limited to, steps of obtaining samples from a subject at different time points in time, monitoring the course of heart failure, obtaining antibodies to determine the stage of heart failure, and measuring miRNA and / or NT-proBNP levels in the subject (or the sample obtained from the subject).

[0146] In some cases, based on the current cohort described in the Experiments section below, biomarker panels containing multiple miRNAs, or biomarker panels containing multiple miRNAs and BNP / NT-proBNP, can be developed. The calculation of predictive scores can be optimized using methods known in this field, such as linear SVM models. In some cases, biomarker panels consisting of varying numbers of miRNA targets can be optimized using SFFS and SVM. The AUC was then optimized for predicting heart failure (Table 26) or classifying heart failure subtypes (Table 27). Examples of formulas, cutoff values, and panel performance are shown in the table.

[0147] [Table 20]

[0148] In Table 26, the symbol "*" represents "×", i.e., the multiplication symbol; "-" represents a negative value; "+" represents a positive value; and "log2(BNP)" represents the base-2 logarithm of BNP expression. The second column of Table 26 also shows examples of formulas for calculating the scores used in the methods used herein. In those formulas, the unit of measurement for microRNA is copies / ml plasma, and for NT-proBNP, it is pg / ml plasma. As will be apparent to those skilled in the art, the coefficients and cutoff values ​​in the formulas will need to be adjusted to suit differences in the detection system used for measurement and / or differences in the units used to express microRNA expression levels and BNP levels / types. It is assumed that the adjustment of the formulas will not exceed the capabilities of the average person skilled in the art.

[0149] In another aspect, a method is provided for determining the risk of progression of heart failure in a subject, or whether the subject has heart failure, which includes (a) measuring the level of a selected panel of miRNAs listed in Table 26 in a sample taken from the subject; and (b) assigning a score based on the level of miRNA measured in step (a) to predict the likelihood of progression of heart failure in the subject, or whether the subject has heart failure. In one example, the score is calculated based on a formula listed in Table 26. In one example, when a biomarker panel of two miRNAs is required, the method can detect and measure the level of the miRNAs listed in Table 26 as “panel of two miRNAs”. In some examples, when a biomarker panel of three miRNAs is required, the method can detect and measure the level of the miRNAs listed in Table 26 as “panel of three miRNAs”. In some examples, when a biomarker panel of four miRNAs is required, the method can detect and measure the level of the miRNAs listed in Table 26 as “panel of four miRNAs”. In some cases, when a biomarker panel of five miRNAs is required, this method can detect and measure the levels of the miRNAs listed in Table 26 as the "panel of five miRNAs". In some cases, when a biomarker panel of six miRNAs is required, this method can detect and measure the levels of the miRNAs listed in Table 26 as the "panel of six miRNAs". In some cases, when a biomarker panel of seven miRNAs is required, this method can detect and measure the levels of the miRNAs listed in Table 26 as the "panel of seven miRNAs". In some cases, when a biomarker panel of eight miRNAs is required, this method can detect and measure the levels of the miRNAs listed in Table 26 as the "panel of eight miRNAs". In some cases, this method can be carried out with an additional step of detecting and measuring the levels of NT-proBNP in the sample.

[0150] In some cases, to predict heart failure, the predictive score (which is thought to be one per subject) is a number indicating the likelihood that the subject has heart failure. In some cases, the results described herein (i.e., the likelihood or prediction of diagnosis) can be found in Equation 3. If the value is greater than a predetermined cutoff value, the subject is diagnosed or predicted to have heart failure. If the value is less than a predetermined cutoff value, the subject is diagnosed or predicted not to have heart failure. Equation 3 is as follows:

number

[0151] [Table 21]

[0152] In Table 27, the symbol "*" represents "×", i.e., the multiplication symbol; "-" represents a negative value; "+" represents a positive value; and "log2(BNP)" represents the base-2 logarithm of BNP expression. The second column of Table 27 also shows examples of formulas for calculating the scores used in the methods used herein. In these formulas, the unit of measurement for microRNA is copies / ml plasma, and for NT-proBNP, it is pg / ml plasma. As will be apparent to those skilled in the art, the coefficients and cutoffs in the formulas will need to be adjusted to suit differences in the detection system used for measurement and / or differences in the units used to express microRNA expression levels and BNP levels / types. It is assumed that the adjustment of the formulas will not exceed the capabilities of the average person skilled in the art.

[0153] In yet another aspect, a method is provided for determining the likelihood that a subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), comprising the steps of (a) measuring the levels of at least two miRNAs listed in Table 27 in a sample taken from the subject; and (b) assigning a score based on the levels of miRNAs measured in step (a) to predict the likelihood that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF). In one example, the score is calculated based on the formula listed in Table 27. In one example, when a biomarker panel of two miRNAs is required, the method can detect and measure the levels of the miRNAs listed in Table 27 as “panel of two miRNAs”. In some examples, when a biomarker panel of three miRNAs is required, the method can detect and measure the levels of the miRNAs listed in Table 27 as “panel of three miRNAs”. In some cases, when a biomarker panel of four miRNAs is required, this method can detect and measure the levels of the miRNAs listed in Table 27 as the "panel of four miRNAs." In some cases, when a biomarker panel of five miRNAs is required, this method can detect and measure the levels of the miRNAs listed in Table 27 as the "panel of five miRNAs." In some cases, when a biomarker panel of six miRNAs is required, this method can detect and measure the levels of the miRNAs listed in Table 27 as the "panel of six miRNAs." In some cases, this method can be carried out with an additional step of detecting and measuring the levels of NT-proBNP in the sample.

[0154] In some cases, to classify between HFREF and HFPEF subtypes, a predictive score (presumably one per subject) can be a number that indicates the likelihood that a heart failure patient has HFPEF subtype heart failure. In some cases, the diagnostic outcome can be found in Equation 4. If the value is greater than a predetermined cutoff value, the heart failure subject is diagnosed (or predicted) to have HFPEF subtype heart failure. If the value is less than a predetermined cutoff value, the subject is diagnosed (or predicted) to have HFREF subtype.

number

[0155] For example, the methods described herein can be implemented as an apparatus capable of (or configured to) perform all (or some) of the steps described herein. Thus, for example, the present disclosure provides an apparatus suitable for (or capable of) performing the methods described herein.

[0156] In another aspect, kits are provided for use (or intended to be used, or when used) in any of the methods described herein. For example, a kit may include reagents to examine the expression of at least one gene listed in Table 9, or at least one gene listed in Table 14, or at least one gene listed in Table 15, or at least two genes listed in Table 16, or at least two genes listed in Table 17, or at least two genes listed in Table 18, or at least two genes listed in Table 19, or at least one gene listed in Table 20, or at least one gene listed in Table 21, or at least one gene listed in Table 22, or at least one gene listed in Table 23, or at least one gene listed in Table 24, or at least one gene listed in Table 25.

[0157] In some cases, the reagent may include probes, primers, or primer sets that are suitable for, or capable of, confirming the expression of at least one gene listed in Table 9, or at least one gene listed in Table 14, or at least one gene listed in Table 15, or at least two genes listed in Table 16, or at least two genes listed in Table 17, or at least two genes listed in Table 18, or at least two genes listed in Table 19, or at least one gene listed in Table 20, or at least one gene listed in Table 21, or at least one gene listed in Table 22, or at least one gene listed in Table 23, or at least one gene listed in Table 24, or at least one gene listed in Table 25.

[0158] In some cases, the kit may further include reagents for determining the levels of brain natriuretic peptide (BNP) and / or the N-terminal prohormone of brain natriuretic peptide (NT-proBNP).

[0159] In some examples, the methods described herein may further include the step of treating a subject predicted to have (or diagnosed with) heart failure or a subtype of heart failure with at least one therapeutic agent for treating heart failure (or a subtype of heart failure). In some examples, the methods may further include therapeutic agents known to alleviate and / or reduce the symptoms of heart failure. In some examples, the methods described herein may further include the administration of drugs (but not limited to) various classes of drugs that have been shown to improve the prognosis of heart failure. Non-limited examples of drugs include ACI / ARBs, angiotensin receptor blockers, loop / thiazide diuretics, beta-blockers, inorganic corticosteroid antagonists, aspirin or Plavix, statins, digoxin, warfarin, nitrates, calcium channel blockers, spironolactone, fibrates, antidiabetic agents, hydralazine, iron supplements, anticoagulants, antiplatelet agents, etc.

[0160] The present invention, as illustrated herein by example, can be successfully implemented without any elements or limitations not specifically disclosed herein. Therefore, terms such as “equipped,” “included,” and “contained” should be read broadly and without limitation. Furthermore, the terms and expressions used herein are for illustrative purposes only, not limiting purposes; therefore, when using such terms and expressions, there is no intention to exclude anything equivalent to or related to the features presented and described herein. However, it will be found that various modifications are possible within the scope of the present invention as described in the claims. Thus, while the present invention has been specifically disclosed by preferred embodiments and optional features, those skilled in the art can utilize modifications and variations of the invention embodied in the preferred embodiments disclosed herein, and such modifications and variations will be considered to fall within the scope of the present invention.

[0161] The present invention has been described broadly and generally in this specification. Each of the narrower species and subgeneral groupings included in the general disclosure also constitutes part of the present invention. This includes the general description of the present invention, except where there is a negative limitation that certain materials are excluded from a genus, regardless of whether those excluded materials are specifically referred to herein.

[0162] Other embodiments are included in the scope of the claims and non-limiting examples below. In addition, if features or aspects of the present invention are described by a group of Markush members, it will be understood by those skilled in the art that the present invention is also described by individual members or parts of members of the group of Markush members. [Examples]

[0163] method Preliminary analysis (sample collection and miRNA extraction): Plasma samples were frozen and stored at -80°C until use. Total RNA was isolated from each 200 μl plasma sample using the well-established TRI reagent (Sigma-Aldrich, registered trademark) according to the manufacturer's protocol. Plasma contains trace amounts of RNA. To reduce RNA loss and monitor extraction efficiency, a rationally designed isolation enhancer (MS2) and spike-in control RNA (MiRXES®) were added to the samples before isolation.

[0164] RT-qPCR: Isolated total RNA and synthetic RNA standards were converted to cDNA in an optimized multiplex reverse transcription reaction with a second set of spike-in control RNA to detect the presence of inhibitors and monitor the efficiency of RT-qPCR. Reverse transcription was performed using Improm II (Promega, Inc.) reverse transcriptase according to the manufacturer's instructions. Next, multiplex amplification was performed on the synthesized cDNA, and quantification was performed using a Sybr Green-based singleplex qPCR assay (MIQE compliant) (MiRXES®). The qPCR reaction was performed using the Applied Biosystems® Vii 7 384 real-time PCR system or the Bio-rad® CFX384 Touch real-time PCR detection system. The overall and detailed workflow for miRNA RT-qPCR measurement is summarized in Figure 4.

[0165] Data Processing: The absolute copy number of target miRNAs in each sample was determined by processing the values ​​of the living cycle against a threshold (Ct) and interpolating them onto a synthetic miRNA standard curve. Technical variability introduced during RNA isolation and the RT-qPCR process was normalized using spike-in control RNA. To analyze single miRNAs, biological variability was further normalized using a group of identified endogenous reference miRNAs that were stably expressed in all control and disease samples.

[0166] result I. Characteristics of Research Participants A well-designed clinical study (case-control study) was conducted to ensure accurate identification of biomarkers for searching for chronic heart failure (HF). This study used a total of 338 patients with chronic heart failure (180 HFREF and 158 HFPEF) from the Singapore population and compared them to 208 non-heart failure subjects as race, sex, and age-matched controls. Patients with heart failure were recruited from the Singapore Heart Failure Outcomes and Phenotypes (SHOP) study

[55] . This included patients with a primary diagnosis of acute decompensated heart failure (ADHF) or patients who visited a hospital for heart failure management because ADHF had appeared within the past six months. Controls without a history of apparent coronary artery disease or heart failure were recruited through the ongoing Singapore Longitudinal Epidemiological Aging Study (SLAS)

[56] . All patients and controls underwent detailed clinical examinations, including comprehensive Doppler ultrasound cardiac imaging, to confirm the presence (or absence) of heart failure in a clinical setting. LVEF was assessed using the disk biplane method, as recommended by the American Society of Cardiology. Patients with confirmed heart failure and an LVEF of 50% or higher were classified as HFPEF, while those with an LVEF of 40% or lower were classified as HFREF. Patients with an EF of 40%–50% were excluded. Evaluations including blood plasma samples were intentionally performed when patients had previously received treatment (typically for 3–5 days), their symptoms had improved, and the clinical physical signs of heart failure had resolved, making discharge appropriate. This ensured accurate evaluation of the performance of markers in the therapeutic or "chronic" phase of heart failure. Clinical characteristics and demographic information are shown in Table 2. All plasma samples were stored at -80°C until use.

[0167] [Table 22]

[0168] [Table 23]

[0169] [Table 24]

[0170] Table 25

[0171] Table 26

[0172] Table 27

[0173] Table 28

[0174] Table 29

[0175] Table 30

[0176] Table 31

[0177] Table 32

[0178] Table 33

[0179] Table 34

[0180] In all samples, plasma NT-proBNP was measured using an automated Cobas e411 analyzer (Roche Diagnostics, Mannheim, Germany) according to the manufacturer's instructions, employing an electro-chemiluminescent immunoassay (Elecsys proBNP II assay). A preliminary investigation of the distribution in the control, HFREF, and HFPEF groups (Figure 2, A-C) revealed that NT-proBNP levels were positively skewed in all groups (skewness / distortion > 2). Since the statistical methods to be applied require an unskewed distribution (Student's t-distribution or logistic distribution), the natural logarithm of NT-proBNP was calculated to generate a new variable, ln_NT-proBNP. The skewness of this variable was close to zero (Figure 2, D-F). ln_NT-proBNP was used in all analyses involving NT-proBNP.

[0181] The characteristics of the subject group are summarized in Table 3.

[0182] [Table 35]

[0183] In addition to demographic variables including age, race, and sex, critical clinical variables for HF, such as LVEF, ln_NT-proBNP, body mass index (BMI), atrial fibrillation or atrial flutter (AF), hypertension, and diabetes, were recorded. While HFPEF patients had a similar mean left ventricular ejection fraction (LVEF) of 60.7±5.9 to healthy control subjects (64.0±3.7), as expected, by selecting and assigning patients, HFREF patients clearly had a lower LVEF (25.9±7.7). Numerical variables were compared between controls and HF (C vs. HF, Table 3), and between HFPEF and HFREF (HFPEF vs. HFREF, Table 3), using Student's t-tests, and classification variables were compared using chi-square tests. Generally, HF patients were older and had a higher frequency of hypertension, AF, and diabetes compared to controls. HFREF patients and HFPEF patients differed in the distribution of sex, age, BMI, hypertension, and AF. In discovering miRNA biomarkers for HF detection and HF subtype classification using multivariate logistic regression, we considered all of these variables with different distributions.

[0184] ln_NT-proBNP was lower in HFPEF than in HFREF, and some results were lower than the ESC-enhanced NT-proBNP cutoff (<125 pg / ml) for HF diagnosis in a non-acute setting

[57] . The decline in NT-proBNP test performance was particularly pronounced in HFPEF (Figure 3, A). The performance of ln_NT-proBNP as a biomarker for HF diagnosis was examined by ROC analysis. In this study, ln_NT-proBNP had an AUC (area below the ROC curve) of 0.962 for the diagnosis of HF as a whole. It was better at detecting HFREF (AUC=0.985) than at detecting HFPEF (AUC=0.935) (Figure 3, B-D). ln_NT-proBNP had an AUC of only 0.706 for the classification of HFREF subtype and HFPEF subtype (Figure 3, C).

[0185] II. Measurement of miRNA Cell-free miRNAs circulating in the blood originate from various organs and blood cells

[58] . Therefore, changes in miRNA levels caused by heart failure may be partially obscured by the presence of the same miRNAs secreted from other sources, possibly due to other stimuli. Thus, determining differences in miRNA expression levels between heart failure and control groups can be challenging. In addition, most cell-free miRNAs are exceptionally scarce in the blood

[59] . Therefore, accurately measuring a large number of miRNA targets from a limited volume of serum / plasma is critical and highly challenging. To facilitate the detection of significantly altered miRNA expression and the identification of a multivariate miRNA biomarker panel for diagnosing heart failure, the inventors of this study chose to perform a well-designed workflow based on qPCR assays, rather than utilizing low-sensitivity or semi-quantitative screening methods (microarrays, sequencing) (Figure 4).

[0186] All qPCR assays (designed by MiRXES® (Singapore)) were repeated at least twice in singleplex for miRNA targets and at least four times for synthetic RNA "spike-in" controls. To ensure accuracy of results in high-throughput qPCR studies, this study designed and established a robust workflow for discovering circulating biomarkers after multiple repetitions (see "Methods" and Figure 4). This novel workflow monitored with various "spike-in" control designs to compensate for technical variability in the isolation, reverse transcription, amplification, and qPCR processes. Each spike-in control was designed in silico to have exceptionally low sequence similarity to known human miRNAs, resulting in non-natural synthetic miRNA mimics (small single-stranded RNAs in the range of 22–24 nucleotides) that minimized cross-hybridization to primers used in the assay. In addition, intentionally splitting the miRNA assays into multiple groups in silico minimized nonspecific amplification and primer-primer interactions. Technical variability was further corrected by constructing a standard curve for absolute copy number interpolation in all measurements using synthetic miRNAs. This highly robust workflow and the use of multiple levels of control enabled this study to identify low levels of circulating miRNA expression. The approach in this study is highly reliable, and data reproducibility is guaranteed.

[0187] Based on previous knowledge regarding highly expressed plasma miRNAs, 203 miRNA targets were selected for this study, and their expression levels were quantitatively measured in a total of 546 plasma samples (HF and control) using a highly sensitive qPCR assay (designed by MiRXES® (Singapore)).

[0188] In this experimental design, total RNA, including miRNA, was extracted from 200 μl of plasma. The extracted RNA was reverse transcribed and amplified by touch-down amplification to increase the amount of cDNA without altering the miRNA expression level (Figure 4). The increased cDNA was then diluted for qPCR measurement. Simple calculations based on the effect of dilution revealed that miRNA expressed at levels below 500 copies / ml in serum would be quantified at a level close to the detection limit of the singleplex qPCR assay (below 10 copies / well). At such concentrations, measurement would be extremely difficult due to technical limitations (errors during pipetting and qPCR reaction). Therefore, miRNA expressed at concentrations below 500 copies / ml was excluded from the analysis and considered undetectable.

[0189] Approximately 70% (n=137) of all miRNAs examined were found to be highly expressed across all samples. These 137 miRNAs were detected in over 90% of the samples (expression level ≥ 500 copies / ml; Table 4). The inventors of this study detected more miRNAs previously not reported in relation to heart failure by comparing them with publicly available data (Table 1). This highlights the importance of using careful and well-controlled experimental design.

[0190] [Table 36]

[0191] [Table 37]

[0192] [Table 38]

[0193] [Table 39]

[0194] III. miRNA biomarkers First, we looked for targets among all the miRNAs measured that were detectable only in heart failure samples and not in control samples. If there were miRNAs specifically secreted by the myocardium of heart failure patients, they would be ideal biomarkers for detecting this disease. Since various organs and / or cells (including myocardium) are known to contribute to miRNAs in the circulatory system, it was not surprising that these miRNAs might already be present in the plasma of healthy individuals and heart failure patients. However, differences in the expression of these miRNAs in plasma could still serve as useful biomarkers during the progression of heart failure.

[0195] A whole-group analysis (principal component analysis, PCA) without a dependent variable was initially performed on the expression levels of all plasma miRNAs (137, Table 4) detected in a total of 546 samples. The first 15 principal components (PCs) with eigenvalues ​​greater than 0.7 were selected and further analyzed, and together these explained 85% of the variance (Figure 5, A). To investigate the difference between controls and heart failure subjects, the AUC was calculated for each of the selected PCs with the aim of classifying these two groups (Figure 5, B). Many PCs were found to have AUCs much larger than 0.5, and even the second PC had an AUC of 0.79. This indicates that the difference between these two groups contributed significantly to the total variance of the miRNA expression profile. Because the variability between control subjects and heart failure subjects was found in multiple dimensions (PC), it was not possible to display all the information based on a single miRNA. Therefore, a multivariate assay involving multiple miRNAs was necessary for optimal classification. Similarly, many PCs had AUCs significantly larger than 0.5 for classifying them into one of two heart failure subtypes (HFREF or HFPEF) (including the first PC (AUC=0.6)), but their AUCs were smaller than the AUC for detecting heart failure (Figure 5, C). Therefore, a multivariate assay capturing information in multiple dimensions was necessary to also classify HFREF and HFPEF.

[0196] When two groups of subjects (C and heart failure (HF)) were plotted on a space defined by two major PCs with discriminative power for HF detection, the two groups were found to be located separately (Figure 6, A). The separation of the HFREF group and the HFPEF group (Figure 6, B) was clearer. Overall analysis revealed that control subjects, HFREF subjects, and HFPEF subjects could be separated based on their miRNA profiles. However, using only one or two dimensions was not statistically robust for classification.

[0197] A key step in identifying biomarkers is to directly compare the expression levels of each miRNA between normal and diseased states, and between disease subtypes. Student's t-test was used for univariate comparisons to assess the significance of group differences for individual miRNAs, and multivariate logistic regression was used to adjust for confounding factors (including age, sex, BMI, AF, hypertension, and diabetes). Bonferroni multiple comparison procedures were used to adjust for false detection rate (FDR) estimation for all p-values ​​

[60] . miRNAs with p-values ​​less than 0.01 were considered significant in this study.

[0198] Next, the expression of 137 plasma miRNAs was compared between A) control (healthy) and heart failure (individual subtypes or combinations of both subtypes), and B) two subtypes of heart failure (i.e., HFREF and HFPEF).

[0199] A) Identification of miRNAs that differ in expression between non-HF control subjects and HF patients.

[0200] Plasma samples were collected from patients clinically identified with one of the following heart failure subtypes (HFREF or HFPEF) and compared to plasma from healthy, non-heart failure donors.

[0201] Initially, a univariate analysis (Student's t-test) was performed to compare the two subtypes, revealing that 94 miRNAs were significantly altered in heart failure patients compared to controls (p-value after FDR < 0.01) (Figure 7, A). Further examination of the two subtypes separately revealed that 82 and 94 miRNAs were significantly altered in HFREF and HFPEF subjects compared to controls, respectively (Figure 7, A). In total, 101 unique miRNAs were identified by univariate analysis, and 75% of them (n=76) were significant for both subtypes (Figure 7A).

[0202] Because control subjects were recruited from the community, their clinical parameters, including the three risk factors for heart failure (AF, hypertension, and diabetes), may not have closely matched those of heart failure patients, as control subjects had fewer of these symptoms. Additionally, there were slight age differences among the analyzed populations. To adjust for these potential confounding factors, multivariate analysis (logistic regression) was performed to examine the significance of miRNAs selected by univariate analysis. In total, 86 out of 101 miRNAs remained significantly different among the surveyed populations after multivariate analysis (Figure 7, B). To detect all heart failures compared to controls, 75 out of 94 miRNAs (Table 5) were found to be significant in multivariate analysis (p-value < 0.01 after FDR). In contrast, 52 out of 82 (Table 6) were significant for detecting HFREF compared to controls, and 68 out of 94 (Table 7) were significant for detecting HFPEF compared to controls (Figure 7B). After multivariate analysis, 36 miRNAs remained significantly different between the control and both heart failure subtypes, 16 differed only between the control and the HFREF subtype, and 32 differed only between the control and the HFPEF subtype (Figure 7, B). Multivariate analysis revealed that many miRNAs differed between the control and only one of the two heart failure subtypes. This suggests that there are true differences between the two subtypes in terms of miRNA expression.

[0203] Table 40

[0204] Table 41

[0205] Table 42

[0206] Table 43

[0207] Table 44

[0208] Table 45

[0209] Numerous miRNAs have been previously reported to be upregulated or downregulated in HF (Table 1). Interestingly, the miRNAs found to have different expression in this study were substantially different from those reported. MiRNAs that were significant in both univariate and multivariate analyses are listed in Table 5 (C vs. heart failure (HF)), Table 6 (C vs. HFREF), and Table 7 (C vs. HFPEF). In the three comparisons, 37, 25, and 33 miRNAs were found to be upregulated, and 38, 27, and 35 miRNAs were found to be downregulated, respectively. The number of miRNAs whose expression was confirmed to be different by qPCR (101 in univariate analysis, and 86 in both univariate and multivariate analyses) was considerably higher than the number previously reported (Table 8, 47 in total). Each of these 86 miRNAs, or any combination thereof, can function as a biomarker for diagnosing heart failure, or as one element of a biomarker panel (multivariate exponential assay).

[0210] [Table 46]

[0211] A total of 47 different miRNAs have been reported in the literature (Table 1). There are conflicting observations regarding the direction of change of hsa-miR-210 in heart failure patients (Table 8). In this study, 22 of the other 46 reported miRNAs were either undetectable or below the detection limit (NA in Table 8), leaving 24 miRNAs available for comparison. Comparing the results of this study (univariate analysis with p-value < 0.01 after FDR) with those of 24 previously reported miRNAs, this study found that only four of those previously reported miRNAs (hsa-miR-423-5p, hsa-miR-30a-5p, hsa-miR-22-3p, hsa-miR-21-5p) were consistently upregulated, while four (hsa-miR-103a-39, hsa-miR-30b-5p, hsa-miR-191-5, hsa-miR-150-5p) were consistently downregulated (Table 8). Interestingly, the direction of change in the eight miRNAs whose regulatory state changed was the opposite of what had been previously reported, while seven remained unchanged (Table 8). Therefore, this study could not confirm the majority of miRNAs that have been previously reported to have different regulatory states in heart failure. Conversely, this study identified more than 70 novel miRNAs that have not been previously reported but could potentially serve as biomarkers for HF detection.

[0212] NT-proBNP / BNP is the most well-studied heart failure biomarker and has demonstrated the best clinical performance to date. Therefore, this study aimed to investigate whether these significantly regulated miRNAs can provide additional information to NT-proBNP. The enhancement of heart failure detection by NT-proBNP by miRNAs was examined using logistic regression adjusted for age, AF, hypertension, and diabetes (p-value, ln_BNP, Tables 5-7). Using a criterion of a p-value less than 0.01 after FDR adjustment, 55 miRNAs (p-value, ln_BNP, Table 7) were found to provide complementary information to ln_NT-proBNP for the detection of HFPEF, but not for the detection of HFREF (p-value, ln_BNP, Table 6). NT-proBNP used alone clearly showed superior diagnostic performance for detecting HFREF (AUC=0.985, Figure 3D) compared to HFPEF (AUC=0.935, Figure 3E). In multivariate assays, combining one or more of these 55 miRNAs with ln_NT-proBNP may improve the detection of HFPEF.

[0213] The AUC values ​​for the most upregulated miRNA (hsa-let-7d-3p, Figure 8, A) and the most downregulated miRNA (hsa-miR-454-3p, Figure 8, B) in heart failure (both subtypes) were 0.78 and 0.85, respectively. Neither miRNA has been previously reported to be useful in detecting heart failure. While the diagnostic power of a single miRNA may not be clinically useful, combining multiple miRNAs in a multivariate manner may improve the performance of heart failure diagnosis.

[0214] B] Identification of miRNAs whose expression differs between HFREF and HFPEF

[0215] Univariate analysis (Student's t-test) revealed that 40 miRNAs differed significantly between HFREF and HFPEF subjects (p-value < 0.01 after FDR), with 10 miRNAs having higher expression levels in HFPEF than in HFREF, and 30 miRNAs having higher expression levels in HFREF than in HFPEF (Table 9).

[0216] Underlying clinical characteristics are expected to differ between the two heart failure subtypes (Table 3). HFPEF patients were more likely to be female, have a higher BMI, be older, and have AF or hypertension compared to HFREF patients. In multivariate analysis (logistic regression) adjusted for these characteristics, only 18 of the 40 miRNAs remained significant (p-value < 0.01 after FDR) (p-value, logistic regression, Table 9). However, since the differences between the two subtypes were due to the natural onset and characterization of the disease rather than bias in the sample selected for this study, all 40 miRNAs in Table 9 (univariate analysis) may be useful for classifying heart failure subtypes.

[0217] The most upregulated miRNA (hsa-miR-223-59, Figure 10, A) and the most downregulated miRNA (hsa-miR-185-5p, Figure 10, B) among all heart failures had moderate AUC values ​​for distinguishing heart failure from controls, at only 0.68 and 0.69, respectively. This is the first report to classify heart failure patients into two clinically important subtypes using circulating cell-free miRNAs from blood (plasma / serum). Combining multiple miRNAs in a multivariate exponential assay will increase the diagnostic power of the subtype classification.

[0218] Reflecting the fact that the degree of dysregulation varies between the two heart failure subtypes, most of the miRNAs that were differentially expressed between HFREF and HFPEF (38 out of 40 in univariate analysis and 17 out of 18 in multivariate analysis) were also found to be different from controls (Figure 11). To further examine the 38 overlapping miRNAs that were found to change in either HF subtype and between the two subtypes in univariate analysis (Figure 9, A), we classified those miRNAs into six groups based on the relationships of their expression levels in three subject groups: control, HFREF, and HFPEF. If the p-value (FDR) for comparison between two groups was greater than 0.01, the relationship was defined as equal (denoted by "="), and if the p-value (FDR) was less than 0.01, the relationship was defined by the direction of change (denoted by greater than ">" or less than "<").

[0219] Gradual changes from control to HFREF and then to HFPEF were found for most miRNAs. Twenty-one miRNAs gradually decreased (C > HFREF > HFPEF, Figure 11), and five miRNAs gradually increased (C < HFREF < HFPEF, Figure 11). Also, five miRNAs were found to be lower only in the HFPEF subtype (C = HFREF > HFPEF, Figure 11), and two were found to be higher only in the HFPEF subtype (C = HFREF < HFPEF, Figure 11), whereas there was no difference between them and controls. Compared to controls, only three miRNAs had levels that were clearly different between HFREF and HFPEF subtypes (C < HFPEF < HFREF or C = HFPEF > HFREF or C = HFPEF < HFREF, Figure 11). Different from LVEF and NT-proBNP, HFPEF had a miRNA profile that was clearly different from the HFREF subtype compared to healthy controls. This suggests that miRNAs may complement NT-proBNP to better discriminate HFPEF.

[0220] From the analysis of all detectable miRNAs, it was revealed that many were positively correlated with each other (Pearson correlation coefficient > 0.5, Figure 12), especially that miRNAs which changed in HF patients and were different between two heart failure subtypes had positive correlations (miRNAs are shown in black towards the right on the x-axis, Figure 12). The changes in miRNA levels in plasma are due to heart failure (HFREF and / or HFPEF). These observations indicate that many miRNA pairs were similarly regulated among all subjects. As a result, a panel of miRNAs could be constructed by systematically optimizing the diagnostic performance by replacing one or more specific miRNAs with another miRNA. All significantly changed miRNAs were extremely important for the development of a multivariate index diagnostic assay for heart failure detection or heart failure subtype classification.

[0221] IV. Plasma miRNAs as Prognostic Markers

[0222] Samples were collected from heart failure patients who, at the index admission when recruited into the SHOP cohort study, were considered appropriate for discharge after 3 - 5 days of treatment when symptoms improved and physical signs of heart failure disappeared clinically. By doing so, it was ensured that in this study the performance of the markers was related to HF in the quasi-acute or "chronic" phase. In this study, the prognostic prediction performance of circulating miRNAs was evaluated for mortality and readmission due to heart failure. 327 heart failure patients (176 with HFREF and 151 with HFPEF) were followed over a 2-year period (Table 10), during which 49 died (15%).

[0223]

Table 47

[0224]

Table 48

[0225]

Table 49

[0226] [Table 50]

[0227] [Table 51]

[0228] [Table 52]

[0229] [Table 53]

[0230] [Table 54]

[0231] Of all cases examined, 115 were readmitted due to heart failure during follow-up (Table 10), and 49 died. For combinations of all-cause mortality and / or re-hospitalization due to decompensated heart failure, miRNAs were evaluated as potential markers for both observed (all-cause survival) OS and event-free survival (EFS).

[0232] Table 11 summarizes the antiheart failure medications prescribed to the participants for study. Comparing the treatment of HFREF and HFPEF, we found that prescription frequencies differed for half of the relevant medications (Figure 13). Classes of medications proven to improve the prognosis of HFREF (ACEIs / ARBs, β-blockers, inorganic corticosteroids) were more commonly prescribed to HFREF patients than to HFPEF patients. Treatment was carried out according to current clinical practice, and the treatment was included as a clinical variable for analyzing prognostic markers.

[0233] Table 55

[0234] Table 56

[0235] Table 57

[0236] Table 58

[0237] Table 59

[0238] Table 60

[0239] Table 61

[0240] Table 62

[0241] Table 63

[0242] Cox proportional hazards (CoxPH) modeling was used in survival analysis, and explanatory variables were analyzed either individually (univariate analysis) or simultaneously (multivariate analysis) using the same model. To better compare different hazard ratios (HRs), all normally distributed variables (including miRNA expression levels (log2 scale)) and clinical variables (e.g., BMI, ln_NT-proBNP, LVEF, age, and multivariate scores resulting from combinations of numerous variables) were scaled to 1 standard deviation. The hazard ratios (HRs) were then used as indicators of the prognostic predictive power of these variables. A p-value < 0.05 was considered statistically significant. Patients were classified into high-risk and low-risk groups based on the presence or absence of a classification variable and whether they were above or below the median of a normally distributed continuous variable. Kaplan-Meier plots (KM plots) were used to visualize the time course of survival for different risk groups, and comparisons between curves were tested using log-rank tests. We also compared intergroup survival at day 750 (OS750) and / or EFS at day 750 (EFS750).

[0243] First, all clinical variables were evaluated for predicting overall survival (OS). In univariate analysis, five variables (age, hypertension, ln_NT-proBNP, nitrate, and hydralazine) were found to be positively associated with the risk of death, and two variables (BMI and β-blockers) were found to be negatively associated with the risk of death (Table 12). Interestingly, there was no difference in overall survival between HFREF patients and HFPEF patients. KM plots of the subject groups defined by these significant parameters are shown in Figure 14A, and OS750 is shown in Figure 14B. All parameters could define high-risk and low-risk groups, with ln_NT-proBNP being the most significant (p-value = 7.2 × 10⁻⁶). -7HR=2.36 (95%CI:1.69~3.30). Based on ln_NT-proBNP levels, the OS750 rate was 92.4% in the low-risk group, compared to only 66.0% in the high-risk group. In a multivariate analysis including all clinical variables, six variables (sex, hypertension, BMI, ln_NT-proBNP, β-blockers, and warfarin) were found to be significant. These six variables were later combined with each of 137 miRNAs in the CoxPH model to identify prognostic miRNA markers for overall survival.

[0244] [Table 64]

[0245] A similar analysis was performed on event-free survival (EFS), and seven variables (AF, hypertension, diabetes, age, ln_NT-proBNP, nitrate, and hydralazine) were found to be positively associated with the risk of recurrent hospitalization due to decompensated heart failure in univariate analysis. KM plots of the subject groups defined by these significant parameters are shown in Figure 15A, and EFS750 is shown in Figure 15B. Here again, there was no difference in event-free survival between HFREF and HFPEF, and ln_NT-proBNP was the most significant (p-value = 1.5 × 10⁻⁶). -9 HR=1.79 (95% CI: 1.42~2.17). ln_NT-proBNP levels below the median were correlated with EFS750 with 65.1%, while levels above the median were correlated with EFS750 with only 34.1%. Multivariate analysis revealed that only two variables (diabetes and ln_NT-proBNP) were significant. These variables were then combined with each of 137 miRNAs to identify prognostic miRNA markers for event-free survival.

[0246] [Table 65]

[0247] To identify miRNA biomarkers for predicting overall survival, each of the 137 miRNAs was tested using a univariate CoxPH model and a multivariate CoxPH model including six additional predictive clinical variables. Thirty-seven (37) were significant in univariate analysis, and 29 were significant in multivariate analysis (Table 14). Of the 11 miRNAs found to be significant in univariate analysis, the predictive performance of clinical parameters could not be improved (multivariate analysis), and three miRNAs were significant only when combined with clinical variables (Figure 16, A). The remaining two miRNAs, with the exception of hsa-miR-374b-5p (p-value = 0.25), had p-values ​​less than 0.1 in univariate analysis (Table 14).

[0248] Regarding mortality, the miRNA with the largest hazard ratio (HR) in both univariate analysis (HR=1.90 (95%CI: 1.36~2.65, p=0.00014)) and multivariate analysis (HR=1.79 (95%CI: 1.23~2.59, p=0.0028)) was hsa-miR-503. For hsa-miR-150-5p, the HR was 0.52 (95%CI: 0.40~0.67, p=1.3×10) in univariate analysis. -7 The smallest difference was observed in both single-variate analysis and multivariate analysis (HR=0.59 (95%CI:0.45~0.78, p-value=0.00032)) (Table 14). KM plots of these two miRNAs are shown in Figure 18A. We can observe that the two risk groups are well separated. Based on a single miRNA, the difference between the high-risk and low-risk groups with respect to OS750 was approximately 21.3% (hsa-miR-503) or approximately 17.8% (hsa-miR-150-5p) (Figure 18, B). When six clinical variables were added, the combined score provided a better risk predictor, with differences of 25.3% for hsa-miR-503 + 6 clinical variables and 22.4% for hsa-miR-150-5p + 6 clinical variables (Figure 18, B). Any one or more of the 40 miRNAs (Table 14) could be used as prognostic markers / panels for the risk of death in patients with chronic HF.

[0249] In univariate analysis (p<0.05), 13 miRNAs were found to be significant in predicting event-free survival, with four positively correlated and nine negatively correlated with the risk of rehospitalization due to decompensated heart failure after treatment (Table 15). In multivariate analysis with two additional clinical variables added to the CoxPH model, no miRNAs were found to be significant in predicting EFS. However, the miRNA with the strongest positive correlation to EFS in univariate analysis (hsa-miR-331-5p, HR=1.27 (95%CI: 1.09~1.49, p=0.0025)) and the miRNA with the strongest negative correlation (hsa-miR-30e-3p, HR=0.80 (95%CI: 0.69~0.94, p=0.0070)) also showed a certain level of significance in multivariate analysis, with p values ​​of 0.15 and 0.14, respectively (Table 15). KM plots of high-risk and low-risk groups defined by miRNAs with and without additional clinical variables for EFS are shown in Figure 19A, and their EFS750 is shown in Figure 19B. Based on a single miRNA (hsa-miR-331-5p or hsa-miR-30e-3p), approximately 40% of the high-risk group and approximately 60% of the low-risk group had an EFS750, whereas when two clinical variables were added, the figures rose to 33% and 66% (Figure 19, B). Any one or more of the 13 miRNAs (Table 15) could be used as a prognostic marker / panel for the risk of rehospitalization due to decompensated HF in chronic HF patients.

[0250] Fewer miRNAs were identified as predictors of event-free survival (n=13) than of overall survival (n=43), with only three overlapping (Figure 16, B). These results suggest that the mechanisms differ between death and relapse-compensated heart failure. One important point to note is that the definition of event-free survival in this study included a poorly defined clinical variable: hospitalization, which may be subject to patient or physician bias from case to case. Nevertheless, a total of 53 miRNAs proved to be valuable prognostic markers for patients with chronic heart failure.

[0251] Next, these 53 prognostic markers were compared to 101 markers for detecting HF (Figure 17, A) or to 40 markers for classifying heart failure subtypes (Figure 17, B). Some overlap was observed, but still, the majority of prognostic markers were not found in the other two lists. This suggests that a different set of miRNAs, or a combination thereof, should be used to form a multivariate exponential assay for prognosis.

[0252] Multivariate biomarker panel for V.HF detection As mentioned above, a panel consisting of multiple miRNA combinations may offer superior diagnostic capabilities compared to using a single miRNA.

[0253] A key criterion for constructing such a multivariate panel was to include at least one miRNA from a special list for each subtype of heart failure to ensure that all heart failure subgroups were covered. However, there was overlap in the miRNAs that defined two subtypes of heart failure (Figure 7). At the same time, many heart failure-related miRNAs and non-heart failure-related miRNAs were found to be positively correlated (Figure 12). Therefore, selecting the best miRNA combination for diagnosing heart failure is difficult.

[0254] Given the complexity of this task, the inventors of this study decided to identify the miRNA panel with the highest AUC using a sequence-forward floating search algorithm

[53] . They also used a current linear support vector machine (a commonly used and well-established modeling tool for constructing a panel of variables) to help select the miRNA combinations

[54] . From this model, a score based on a linear formula that takes into account the expression level and weight of each member was generated. These linear models were readily applicable to clinical practice.

[0255] A crucial condition for the success of these methods is the availability of high-quality data. Quantitative data of all miRNAs detected in a large number of well-defined clinical samples not only improves the accuracy and precision of the results but also ensures the consistency of the biomarker panel identified for subsequent clinical application using qPCR.

[0256] To ensure the authenticity of the results, hold-out confirmation (two-part cross-validation) was performed multiple times (over 80 times) to examine the performance of the biomarker panel identified based on the discovery set (half of the sample in each run) against an independent set of confirmation samples (the remaining half of the sample in each run). The use of a large number of clinical samples (546) minimized the problem of data overfitting in modeling because only 137 candidates were selected from this sample, while 273 samples were used as the discovery set each time, resulting in a sample-to-candidate ratio greater than 2. During the cross-validation process, samples were matched for subtype, sex, and race. This process was then performed to separately optimize biomarker panels containing 3, 4, 5, 6, 7, 8, 9, or 10 miRNAs.

[0257] A box plot representing the results (AUC of the biomarker panel in both the discovery and confirmation phases) is shown in Figure 20A. The AUC values ​​were very close to each other across different discovery sets (box size < 0.01) and approached 1 (AUC = 1) as the number of miRNAs in the panel increased. When there were four or more miRNAs, the box size (indicating the spread of values) in the confirmation phase was also very small (AUC value less than or equal to 0.01). As expected, the AUC value decreased in the confirmation set for each discovery (AUC between 0.02 and 0.05).

[0258] A more quantitative representation of the results is shown in Figure 20B. While the AUC consistently increased gradually in the discovery phase when the number of miRNAs in the biomarker panel was increased, there was no further significant improvement in the AUC value in the confirmation phase when the number of miRNAs was 9 or more. The difference between the biomarker panel with 6 miRNAs and the biomarker panel with 8 miRNAs was statistically significant, but the improvement in AUC value was less than 0.01. Therefore, a biomarker panel containing 6 or more miRNAs and yielding an AUC value of approximately 0.93 should be useful for detecting heart failure.

[0259] To examine the composition of the multivariate biomarker panel, this study counted the occurrence of miRNAs in all panels containing 6–10 miRNAs, excluding the top 10% and bottom 10% of panels by AUC. This was done to avoid counting biomarkers that might be misidentified due to inaccurate data fitting from small populations during the randomization process in cross-confirmation analysis. By excluding these miRNAs selected by less than 2% of the panels, a total of 51 miRNAs were selected in the discovery process (Table 16), and it was found that the expression of 42 of these was significantly altered in HF (Tables 5–7). The other 9 miRNAs were not altered in heart failure, but their inclusion was found to significantly improve AUC values ​​because 39% of the panels contained at least one of these miRNAs from this list, and the most frequently selected miRNA (hsa-miR-10b-5p) was present in 35% of the panels. Without direct and quantitative measurement of all miRNA targets, these miRNAs would never have been selected in high-throughput screening studies (microarrays, sequencing), and would have been excluded from confirmation by qPCR.

[0260] When comparing the miRNAs selected for multivariate panels with the single miRNAs used as diagnostic markers, they were not always the same. For example, the upregulated top-level miRNA (hsa-let-7d-3p) was not present in the list, and the downregulated top-level miRNA (hsa-miR-454-3p) was used in only 24.2% of the panels. Therefore, simply combining the best single miRNAs identified did not produce an optimal biomarker panel; rather, a panel of miRNAs providing complementary information yielded the best results.

[0261] Not all of these miRNAs were randomly selected, as seven of them were present in over 30% of the panels. However, it was difficult to identify miRNAs that were crucial for a superior biomarker panel. This was because the two most frequently selected miRNAs, hsa-miR-551b-3p and hsa-miR-24-3p, were found in only 59.7% and 57.3% of the panels, respectively. As mentioned above, many of these miRNAs were correlated (Figure 11), allowing them to be substituted or replaced by each other within the biomarker panel. In conclusion, a biomarker panel containing at least six miRNAs from the frequently selected list (Table 16) should be used for the detection of heart failure.

[0262] To compare miRNA biomarkers with NT-proBNP, one miRNA was selected from a biomarker panel of six, and the total miRNA score for all subjects was calculated and plotted against NT-proBNP levels from the same subjects (Figure 21, A). The inventors of this study generally observed a positive correlation, with a Pearson correlation coefficient of 0.61 between miRNA score and ln_NT-proBNP (p-value = 8.2 × 10⁻¹⁰). -56Applying the indicated cutoff for NT-proBNP (125 pg / ml, dotted line), 35 healthy subjects were misclassified as heart failure patients (false positive, FP, NT-proBNP > 125), and 23 heart failure patients had NT-proBNP levels below the cutoff (false negative, FN). As expected, the majority of false negative (FN) subjects were HFPEF subjects (n=20). These false positive (FP) and false negative (FN) subjects for NT-proBNP were selected, and the results were plotted against miRNA scores (Figure 21, B). Based on this separated plot, using zero as the cutoff value allowed for the correct reclassification of the majority of false positive (FP) and false negative (FN) subjects by miRNA scores (dotted line). These results confirmed the hypothesis that miRNA biomarkers provide different information than NT-proBNP. The next step was to look for a multivariate biomarker panel that included both miRNA and NT-proBNP.

[0263] The same biomarker identification process (multiple 2-fold cross-validation) was performed. NT-proBNP was fixed as one of the predictor variables, and a classification mechanism using a support vector machine was constructed using the level of ln_NT-proBNP and the expression level of miRNAs (log2 scale). Since the AUC did not significantly increase when predicting heart failure using 9 or more miRNAs, this process was performed to optimize biomarker panels containing 3, 4, 5, 6, 7, 8, 9, or 10 miRNAs (and NT-proBNP).

[0264] The classification mechanism constructed in the discovery phase approached complete segregation (AUC=1.00) as the number of miRNAs increased. Performance slightly decreased in the confirmation phase (Figure 22, A). Nevertheless, the AUC of the panel containing NT-proBNP in the confirmation phase (mean AUC>0.96) was consistently higher than that of the biomarker panel containing only miRNAs (mean AUC<0.94). Quantitative results (Figure 22, B) showed that the AUC value did not improve significantly further in the confirmation phase when the number of miRNAs was 5 or more, and there was only a slight increase (0.001 AUC) in the biomarker panel with 4-5 miRNAs. Therefore, when combined with NT-proBNP, a biomarker panel containing 4 or more miRNAs and giving an AUC value of approximately 0.98 can be used for the detection of heart failure. Combining miRNAs with NT-proBNP significantly improved classification efficiency compared to NT-proBNP alone (AUC=0.962, Figure 22, B).

[0265] The top 10% and bottom 10% of panels by AUC were excluded, and the composition of multivariate biomarker panels containing 3 to 8 miRNAs was examined (Table 17). A total of 49 miRNAs were selected in the discovery process, of which 14 had a presence rate of over 10% (Table 17). Of these, 42 also provided additional information for NT-proBNP (p-value less than 0.01 after FDR in logistic regression). Here again, 46% of the panels contained at least one of the 13 miRNAs found to be insignificant, in addition to NT-proBNP.

[0266] While more than half of the significant miRNAs (Table 17, list of significant miRNAs) were frequently selected when searching for a biomarker panel consisting solely of miRNAs (Table 16, list of significant miRNAs) (Figure 23, A), their ranking by prevalence differed. Some miRNAs frequently selected in combination with NT-proBNP (hsa-miR-17-5p (11.6%) and hsa-miR-25-3p (11.0%)) were not even selected when searching for a miRNA-based biomarker panel (without NT-proBNP). Furthermore, only two miRNAs overlapped between the lists of insignificant miRNAs (Table 16 and Table 17, list of insignificant miRNAs) (Figure 23, B). Taken together, this evidence suggests that a different list of miRNAs should be used in combination with NT-proBNP compared to the list used to construct a biomarker panel consisting solely of miRNAs.

[0267] VI. Multivariate biomarker panels for heart failure subtype classification To identify a multivariate biomarker panel for distinguishing between HFREF and HFPEF, the following attempt was made. Again, full quantitative data for 137 miRNAs from 338 heart failure patients were used. Due to sample size constraints, multiple (over 50) quadrilateral cross-confirmation studies were performed. In each study, all subjects were randomly divided into four equal groups, and a classification mechanism was constructed using three of these groups (discovery groups) to predict the final group (confirmation group). Thus, the discovery phase used 253–254 subjects to ensure that each subgroup (HFREF or HFPEF) was of similar size. This size is similar to the number of candidates (137) selected to minimize overfitting. This process was repeated to separately optimize biomarker panels containing 3, 4, 5, 6, 7, 8, 9, or 10 miRNAs, and biomarker panels containing 2, 3, 4, 5, 6, 7, or 8 miRNA+NT-proBNTs.

[0268] Quantitative results showed that the miRNA-only biomarker panel did not improve AUC values ​​when it contained six or more miRNAs (Figure 24, A). An AUC of approximately 0.76 could be achieved using the miRNA biomarker panel, which was superior to NT-proBNT (AUC=0.706). When all panels with 6-10 miRNAs (excluding the top and bottom 10% of AUC) were counted, 46 miRNAs were frequently selected (less than 2% of the panel), of which 22 were found to be significant in t-tests comparing HFREF and HFPEF, while 24 were not (Table 18). The panel for heart failure subtype classification showed less diversity than the panel for heart failure detection because two of the miRNAs were present in over 80% of the panels (hsa-miR-30a-5p (94.6%) and hsa-miR-181a-3p (83.7%), Table 18).

[0269] Biomarker panels consisting of both miRNAs and NT-proBNT required fewer miRNAs than panels containing only miRNAs. This was because including more than four miRNAs did not improve the AUC value (Figure 24, B). Compared to panels containing only miRNAs, clearer classification was achieved (AUC = 0.82). Here again, miRNAs and NT-proBNTs can provide complementary information for heart failure subtype classification. When the composition of panels containing 5-8 miRNAs plus NT-proBNT was examined, 31 miRNAs were frequently selected (over 2% of the panel), of which 14 were found to be significant in logistic regression combined with ln_NT-proBNT, while 17 were not (Table 19). Two different miRNAs were found in over 80% of the panels (hsa-miR-199b-5p (91.5%) and hsa-miR-191-5p (74.9%)). In both the miRNA-only panel and the miRNA+NT-proBNT panel, the most frequently selected non-significant miRNA was the same (hsa-miR-199b-5p), but a significant difference was found between the remaining lists of significant and non-significant miRNAs in terms of their type and rank.

[0270] [Table 66]

[0271] [Table 67]

[0272] [Table 68]

[0273] [Table 69]

Claims

1. A method for determining whether a subject has or is at risk of developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), wherein this method is a) A step of measuring the level of at least hsa-miR-181a-2-3p in a bodily fluid sample obtained from the subject, b) The process includes determining whether the level is different from that of a control, Compared to the threshold determined from the levels in the control, an increased level of hsa-miR-181a-2-3p indicates that the subject suffers from heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), or is at risk of developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), wherein the method further includes measuring the levels of at least hsa-miR-486-5p and hsa-miR-30a-5p in a body fluid sample obtained from the subject. Here, a decrease in the level of hsa-miR-486-5p compared to the threshold determined from the levels in the control group indicates that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), or is at risk of developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), and The method wherein a decrease in the level of hsa-miR-30a-5p compared to a threshold determined from the level in the control indicates that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), or is at risk of developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF).

2. A method for determining whether a subject has or is at risk of developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), the following: (a) A step of measuring the levels of at least hsa-miR-181a-2-3p, hsa-miR-486-5p and hsa-miR-30a-5p in a body fluid sample obtained from a subject; and (b) A step of determining whether the level is different from that of a control, including a step of predicting the risk of the subject having or developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with preserved left ventricular ejection fraction (HFPEF), Here, an increase in the level of hsa-miR-181a-2-3p compared to the threshold determined from the levels in the control indicates that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), or is at risk of developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF). Here, a decrease in the level of hsa-miR-486-5p compared to the threshold determined from the levels in the control group indicates that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), or is at risk of developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), and The method wherein a decrease in the level of hsa-miR-30a-5p compared to a threshold determined from the level in the control indicates that the subject has heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF), or is at risk of developing heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF).

3. Step (a) further comprises measuring the level of at least one further miRNA in a body fluid sample obtained from the subject, wherein the miRNA is one of the following: hsa-miR-223-5p, hsa-miR-335-5p, hsa-miR-452-5p, hsa-miR-23b-3p, hsa-miR-181b-5p, hsa-miR-146a-5p, hsa-miR-199b-5p, hsa-miR-126-5p, hsa-miR-23a-5p, hsa-miR-185-5p, hsa-miR-20b-5p, hsa-miR-550a-5p, hsa-miR-106a-5p, hsa-let-7b-5p, hsa-miR-93-5p, hsa-miR-20a-5p, hsa-miR-25-3p, hsa-miR-18b-5p, hsa-miR-532-5p, hsa-miR-501-5p, hsa-miR-4732-3p, hsa-miR-144-3p, hsa-miR-192-5p, hsa-miR-17-5p, hsa-miR-363-3p, hsa-miR-103a-3p, hsa-miR-16-5p, hsa-miR-194-5p, hsa-miR-183-5p, hsa-miR-451a, hsa-miR-19b-3p, hsa-miR-106b-3p, hsa-miR-19a-3p, hsa-let-7i-5p, hsa-miR-196b-5p, The method according to claim 1 or 2, wherein the selected molecule is from the group consisting of hsa-miR-500a-5p and hsa-miR-122-5p.

4. The method according to any one of claims 1 to 3, wherein the control is a subject having heart failure with reduced left ventricular ejection fraction (HFREF) or heart failure with maintained left ventricular ejection fraction (HFPEF).

5. The method according to any one of claims 1 to 4, wherein the method comprises determining the expression levels of at least four, at least five, at least six, at least seven, or at least eight miRNAs.

6. The method according to any one of claims 1 to 5, wherein the body fluid is selected from the group consisting of amniotic fluid, milk, bronchial lavage fluid, cerebrospinal fluid, colostrum, interstitial fluid, peritoneal fluid, pleural fluid, saliva, semen, urine, tears, whole blood, plasma, and serum, including their cellular and noncellular components.

7. The method according to any one of claims 1 to 6, wherein the subject is of Asian ethnicity.

8. A kit for use in the method according to any one of claims 1 to 7, comprising a reagent for determining or measuring the expression level of the miRNA described in claim 1, 2, or 3.

Images