Differential gene analysis

Figure 1a shows box plots of 28 microRNA genes expressed locally in the CC and healthy control groups. There were 11 MRGs with p values < 0.05, among which 5 genes were upregulated in the cervical cancer group, whereas 6 genes were downregulated in the cervical cancer group. Figure 1b presents a heatmap of MRGs expression in the two groups. The correlations among MRGs in the cervical cancer group are shown in Fig. 1c, d, where MAP1LC3A, MAP1LC3B, and PINK1 are closely related to MRGs. A PPI network of the 28 MRGs was subsequently constructed (Fig. 1e, f).

Fig. 1

The expression of mitophagy-related genes (MRGs) in cervical cancer was analysed via various approaches (a, b). Additionally, a correlation heatmap was generated to illustrate the relationships among MRGs, specifically in the CC group (c). Furthermore, protein‒protein interaction (PPI) networks were constructed to visualize the interactions among MRGs in CC (d) and to highlight the top 10 hub genes on the basis of MRGs (e). Statistical significance is denoted as follows: *P < 0.05, ** P < 0.01, *** P < 0.001, NS: not significant

LASSO analysis

LASSO regression analysis was performed on 28 MRGs, resulting in a cervical cancer diagnostic risk scoring model consisting of 23 MRGs (see Fig. 2). The ROC curve (AUC = 0.8912).

Fig. 2

A disease model was constructed via LASSO analysis. LASSO coefficients were screened (a). A trajectory diagram of the LASSO variables was generated, where each curve represents the coefficient trajectory of an independent variable (b). Different trajectories correspond to varying LASSO coefficients as lambda changes. An ROC curve was generated for a cervical cancer diagnostic risk score model based on MRGs (c).

Enrichment gene analysis

On the basis of the differentially expressed MRGs, we conducted GOKEGG pathway enrichment analyses (Fig. 3). The GO annotation results indicated that the differentially expressed MRGs were associated primarily with biological processes (BPs), such as macroautophagy and autophagy. KEGG pathway enrichment analysis revealed that these genes were involved in processes such as ferroptosis (Fig. 4). Additionally, gene set enrichment analysis (GSEA) revealed enrichment of neutrophil degranulation via the Pathway Interaction Database (PID). Moreover, the gene set variation analysis (GSVA) results (Fig. 5) revealed enrichment of the GO terms microfibril and amino acid betaine metabolic process.

Fig. 3

GO and KEGG enrichment gene analyses. This included the examination of biological processes (a), cellular components (b), molecular functions (c), and KEGG pathways (d). GO gene ontology, KEGG Kyoto Encyclopedia of Genes and Genomes

Fig. 4

Gene set enrichment analysis (GSEA) revealed differential enrichment of various signalling pathways in the cervical cancer samples (a–l). Significance was determined at a P value < 0.05 for pathway enrichment

Fig. 5

Changes in pathway activity were analysed in patients with cervical cancer via gene set variation analysis (GSVA). A volcano map was created to illustrate the differential GSVA enrichment between normal and cancerous samples (a). A cluster heatmap was generated to display the pathways of GSVA in the two groups (b), and a box plot was utilized to show the enrichment levels of pathways in the two groups (c). GSVA gene set variation analysis

WGCNA nanolysis

Unsupervised clustering was conducted by employing systematic WGCNA for partitioning. The identification and color assignment of these modules were achieved through merged dynamic tree cutting, resulting in a total of 13 distinct modules. Notably, the blue module exhibited a positive correlation with the samples (Fig. 6).

Fig. 6

This study focused on WGCNA. A matrix was formed to depict the relationships among the modules and their characteristics (a, b). WGCNA was then employed to evaluate the correlation ® between external factors such as epileptic or normal conditions (c–h)

MRGs model analysis

Figure 7a displays the top six coexpressed genes. The Venn diagram in Fig. 7b compares the MRGs with the MEturquoise modules. We subsequently constructed a protein‒protein interaction (PPI) network via the STRING database (Fig. 7c) and imported the interactions into Cytoscape software for further analysis. By employing the CytoHubba plug-in within Cytoscape, we identified the top 10 hub genes, as depicted in Fig. 7d. Finally, the Sankey diagram in Fig. 7e illustrates the prediction of lncRNAs and miRNAs.

Fig. 7

The interaction network was constructed via various approaches. An UpSet diagram was generated to illustrate the relationships between the gene coexpression modules and marker genes of interest (a). A Venn diagram was generated to depict the overlap between the MEturquoise modules, marker genes, and differentially expressed marker genes (b–e)

Clustering analysis

In this study, through bioinformatics analysis, we focused on screening cervical cancer susceptibility genes and exploring the construction of an MRGs diagnostic model. In the cluster analysis, we systematically integrated different classifications to reveal the expression patterns of CC susceptibility genes in multiple dimensions. This comprehensive classification method provides a new perspective for understanding the pathogenesis of cervical cancer. Moreover, we investigated the role of RBPs in the development of cervical cancer in depth. By analysing the relationships between RBPs and different classifications, we found that there are close associations between multiple RBPs and cervical cancer susceptibility genes. These findings provide important insight into the regulatory mechanism of RBPs in the development of cervical cancer (Fig. 8).

Fig. 8

The molecular subtypes of cervical cancer were determined via gene expression levels. Unsupervised consensus clustering (a–f) of gene expression data from cervical cancer samples revealed the presence of 2–6 distinct clusters (g–h)

CIBERSORT was utilized to perform immune infiltration analysis

Our study assessed the abundance of immune cells in cervical cancer and normal tissue samples, as illustrated in Fig. 9a, b. The findings revealed a heightened presence of neutrophils (P < 0.01) in patients with cervical cancer; additionally, a correlation heatmap depicted the associations among the hub genes identified in the PPI network (Fig. 9c, d).

Fig. 9

CIBERSORT analysis was performed to evaluate immune infiltration in cervical cancer. Histogram displaying the distribution of 22 immunocyte subgroups in CC samples (a) and an examination of differences in immune infiltration between control and CC samples (b); correlation heatmaps (c, d)

Single-cell sequencing

Figure 10A shows a t-SNE scatter plot, where different colors represent different types of cells in cervical cancer samples, including T cells, B cells, tumor cells, fibroblasts, monocytes, macrophages, neutrophils, and other cells. Tumor cells and T cells occupied significant positions in the plot, indicating their greater abundance in the cervical cancer samples. Figure 10B shows a gene expression dot plot displaying the expression levels of multiple genes in different cell types. Notably, certain genes are expressed at higher levels in specific cell types, for example, certain genes are expressed at higher levels in T cells than in other cell types. Figure 10C presents a gene expression heatmap, illustrating the expression levels of different genes in the cervical cancer samples. The color gradient from light to dark represents the intensity of gene expression, with some genes showing higher expression in tumor cells and lower expression in other cell types. These visualizations collectively reveal the complex patterns of different cell types and gene expression in cervical cancer samples, aiding in a deeper understanding of the biological characteristics of cervical cancer.

Fig. 10

Single-cell sequencing. T-SNE clustering plot of cell types (a); expression dot plot of target genes in different cell types (b); heatmap of target gene expression (c); network plot of interactions between different cells (d); heatmap of signalling patterns in different cell types (e); heatmap of signalling patterns in different cell types (f)

Figure 10D depicts an interaction network among different cell types in cervical cancer samples. Nodes represent different cell types (such as T cells, B cells, and tumor cells), whereas edges represent their interactions. Tumor cells clearly have dense interactions with other cell types, particularly strong connections with T cells and fibroblasts, reflecting intricate cell communication within the tumor microenvironment. Figure 10E displays a heatmap of signalling patterns of different cell types, with a color gradient indicating signal strength. The results revealed that tumor cells exhibit strong signal outputs in multiple signalling pathways (such as the TGFβ, Wnt, and EGF pathways), suggesting the potential role of these pathways in regulating tumor cell functions. Figure 10F shows a heatmap of signalling patterns received by different cell types, with color gradients similarly representing signal strength. Tumor cells receive signals from mainly the TGFβ, Wnt, and EGF pathways, which may play crucial roles in tumor cell growth and survival. These visualizations collectively reveal the intricate signalling networks among cells in cervical cancer samples, offering important insights into cell communication within the tumor microenvironment.

In this study, immune cells in cervical cancer tissue were analysed in detail via single-cell sequencing. The results revealed that immune cells exhibit specific aggregation patterns in the tumor microenvironment, especially in the tumor margin region (Fig. 11). Atlas analysis further revealed the spatial distribution characteristics of different immune cell subpopulations and their interactions with tumor cells (Fig. 12). Cytospectral density analysis revealed that certain immune cell subpopulations dominated high-density regions, which may be related to the immune escape mechanism of tumors (Fig. 13).

Fig. 11

Specific aggregation patterns of immune cells in the tumor microenvironment were analysed by single-cell sequencing

Fig. 12

The spatial distribution characteristics of different immune cell subsets were analysed via single-cell sequencing

Fig. 13

The cell spectral density was analysed via single-cell sequencing

Verification by in vitro cell experiments

Compared with the NC mimic group, the miRNA mimic group presented a significant reduction in the number of mitochondria in Hela cells, resulting in poor fluorescence intensity. This difference was statistically significant (P < 0.01) (Fig. 14A). Additionally, compared with that in the NC mimic group, the red fluorescence in the Hela miRNA mimic group decreased, whereas the intensity of green fluorescence increased. The statistical analysis revealed significant differences in the relative values of red and green fluorescence between the two groups (P < 0.001) (Fig. 14B). Mitochondrial permeability transition was assessed via the use of a calcein AM fluorescent probe. The results revealed a significant decrease in green fluorescence in Hela cells in the miRNA mimic group compared with the NC mimic group, with a statistically significant difference in fluorescence values between the two groups (P < 0.001) (Fig. 14C).

Fig. 14

Effects of miR-431-5p overexpression on mitochondrial function in Hela cells. Mitochondrial number (a), mitochondrial potential (b), and mitochondrial mPTP (c). **P < 0.01, ***P < 0.001 vs. the NC mimic. mPTP mitochondrial permeability transition pore, NC negative control