Identification of ageing-associated gene signatures in heart failure with preserved ejection fraction by integrated bioinformatics analysis and machine learning

Abstract

The incidence of heart failure with preserved ejection fraction (HFpEF) increases with the ageing of populations. This study aimed to explore ageing-associated gene signatures in HFpEF to develop new diagnostic biomarkers and provide new insights into the underlying mechanisms of HFpEF. Mice were subjected to a high-fat diet combined with L-NG-nitroarginine methyl ester (l-NAME) to induce HFpEF, and next-generation sequencing was performed with HFpEF hearts. Additionally, separate datasets were acquired from the Gene Expression Omnibus (GEO) database. The differentially expressed genes (DEGs) were used to identify ageing-related DEGs. Support vector machine, random forest, and least absolute shrinkage and selection operator algorithms were employed to identify potential diagnostic genes from ageing-related DEGs. The diagnostic value was assessed using a nomogram and receiver operating characteristic curve. The gene and related protein expression were verified by reverse transcription PCR and western blotting. The immune cell infiltration in hearts was analysed using the single-sample gene-set enrichment analysis algorithm. The results showed that the merged HFpEF datasets comprised 103 genes, of which 15 ageing-related DEGs were further screened in. The ageing-related DEGs were primarily associated with immune and metabolism regulation. AGTR1a, NR3C1, and PRKAB1 were selected for nomogram construction and machine learning-based diagnostic value, displaying strong diagnostic potential. Additionally, ageing scores were established based on nine key DEGs, revealing noteworthy differences in immune cell infiltration across HFpEF subtypes. In summary, those results highlight the significance of immune dysfunction in HFpEF. Furthermore, ageing-related DEGs might serve as promising prognostic and predictive biomarkers for HFpEF.

Keywords: Ageing; Bioinformatics analysis; HFpEF; Immune dysfunction; Machine learning.

Figure 1

Identification of differentially expressed genes (DEGs) and ageing-related DEGs (ARDEGs). (A–C) The volcano plot of the DEGs in our dataset (heart failure dataset) (A), GSE194151 dataset (B), and GSE184120 dataset (C). (D) Identification of the common DEGs. (E) Identification of the ARDEGs. (F–H) Heatmap of ARDEGs in the heart failure (F), GSE194151 (G), and GSE184120 (H) datasets.

Figure 2

Diagnostic biomarker identification and verification based on ARDEGs via machine learning. (A) The forest plot illustrating ARDEG expression via the logistic regression analysis. (B, C) The number of genes with the lowest error rate (B) and the highest accuracy rate (C) in the SVM model. (D, E) Random forest analysis was conducted to analyse the ARDEGs and extract potential diagnostic biomarkers. (F, G) Biomarker screening via LASSO regression analysis. (H–K) The visible nomogram for diagnosis (H), and the diagnostic value evaluation (I–K). (L) The Venn diagram showing five candidate diagnostic genes identified via SVM, logic-LASSO, and random forest algorithms. SVM, support vector machine; ARDEGs, ageing-related differentially expressed genes; LASSO, least absolute shrinkage and selection operator; ROC, receiver operating characteristic; AUC, the area under the ROC curve; DCA, decision curve analysis.

Figure 3

The expression of five candidate diagnostic genes and the verification of diagnostic specificity and sensitivity. (A–C) The expression of candidate diagnostic genes in heart failure (A), GSE194151 (B), and GSE184120 datasets (C). (D–L) The ROC curve of each candidate gene (Agtr1a, NR3C1, and PRKAB1) in the heart failure (D–F), GSE194151 (G–I), and GSE184120 datasets (J–L). Not significant, p ≥ 0.05; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. HFpEF, heart failure with preserved ejection fraction; ROC, receiver operating characteristic; AUC, the area under the curve.

Figure 4

The mRNA and protein levels of Agtr1, Nr3c1, and Prkab1 in hearts from HFpEF mice were verified. (A–C) The mRNA expression of Agtr1, Nr3c1, and Prkab1 in HFpEF hearts. (D–G) The levels of protein encoded by Agtr1, Nr3c1, and Prkab1 in HFpEF hearts. Not significant, p ≥ 0.05; ∗p < 0.05, ∗∗p < 0.01. HFpEF, heart failure with preserved ejection fraction.

Figure 5

Immune cell infiltration analysis between the low risk score and high risk score groups based on the ARDEG diagnostic model. (A) The proportion of 28 immune cell types in the low risk score group and high risk score group visualized by the bar plot. (B–P) The correlation of T follicular helper cell infiltration and ARDEGs was analyzed via the Spearman algorithm. Not significant, p ≥ 0.05; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. HFpEF, heart failure with preserved ejection fraction; ssGSEA, single-sample gene set enrichment analysis; ARDEG, ageing-related differentially expressed gene.

Figure 6

HFpEF subtype identification based on the ARDEG diagnostic model. (A) The heatmap exhibiting the two HFpEF clusters with k = 2 based on the ARDEGs. (B) Cumulative distribution function (CDF) for k = 2–9. (C) Delta diagram illustrating the variations of the area under the CDF curve for k = 2–9. (D) PCA based on the results of the consensus clustering analysis. (E) ARDEG expressions in two different HFpEF clusters. (F) Volcano plot of the DEGs in two different HFpEF clusters. (G) Venn diagram of nine key DEGs identified by intersecting the DEGs and ARDEGs. (H) The heatmap exhibiting expression of the key DEGs in two different HFpEF clusters. HFpEF, heart failure with preserved ejection fraction; CDF, cumulative distribution function; PCA, principal component analysis; DEGs, differentially expressed genes; ARDEGs, ageing-related DEGs.

Figure 7

Immune cell infiltration analysis between two different HFpEF clusters. (A) The proportion of 28 immune cell types in two different HFpEF clusters (cluster 1 and cluster 2). (B, C) Correlation of 24 immune cell types with a significantly different infiltration abundance in cluster 1 (B) and cluster 2 (C). (D, E) The correlation of 24 immune cell types with a significantly different infiltration abundance and key DEGs in cluster 1 (D) and cluster 2 (E). HFpEF, heart failure with preserved ejection fraction; ssGSEA, single-sample gene set enrichment analysis; DEGs, differentially expressed genes.

Figure 8

Construction of A-scores and the verification of diagnostic specificity and sensitivity. (A) The ROC curve of A-scores for the GSE194151 dataset. (B) The expression of key DEGs in the high and low A-score groups. (C–K) The ROC curve of each key DEG shows the diagnostic value for the HFpEF subtypes based on the A-scores. Not significant, p ≥ 0.05; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. HFpEF, heart failure with preserved ejection fraction; DEGs, differentially expressed genes; A-scores, ageing scores; KM, Kaplan–Meier; ROC, receiver operating characteristic; AUC, area under the ROC curve.

Figure 9

Immune cell infiltration analysis between the high and low A-score groups. (A) The proportion of 28 immune cell types in the high and low A-score groups. (B, C) Correlation of 18 immune cell types with a significantly different infiltration abundance in the low (B) and high (C) A-score groups. (D, E) The correlation of 18 immune cell types with a significantly different infiltration abundance and key DEGs in the high and low A-score groups. Not significant, p ≥ 0.05; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. HFpEF, heart failure with preserved ejection fraction; ssGSEA, single-sample gene set enrichment analysis; A-score, ageing score; DEGs, differentially expressed genes.