3.1 Expression profiles of PRGs in LGGs and their significantly associated miRNAs

The main flow chart of this study is shown in Supplementary Fig. 2. As shown in Table 1, 16 PRGs were screened and borrowed to import the 16 PRGs into the STRING database to obtain the PPIs of PRGs (Fig. 1A), which were mainly enriched in terms of biological processes in response to oxidative stress, and were highly correlated with the signalling pathways, such as Necroptosis, Apoptosis, etc. (Fig. 1B). The expression of PRGs in tumour tissues and normal tissues was observed by IHC staining results (Fig. 1C–R). Spearman correlation analysis screened to obtain 91 miRNAs significantly correlated with PRGs (Fig. 2A), and constructed a network diagram with the relationship between PRGs and miRNAs (Fig. 2B). As a result, it was found that there were three PRGs (PARP1, AIFM1, and GPX4) associated with more than two miRNAs missing strong correlations and were left out of the network. In addition to this, PRGs exhibited complex regulatory relationships with these miRNAs (Fig. 2C).

Table 1 Detailed list of 16 Parthanatos related genesFig. 1

Overview of 16 Parthanatos-Related Genes (PRGs). A PPI network for the 16 PRGs; B Enrichment analysis of the 16 PRGs conducted using the Human Protein Atlas; Expression of (C) PARP1; D AIFM1; E RNF146; F ADPRS; G CYBB; H SIRT1; I NAT10; J HK1; K NCF1; L NAMPT; M GPX4; N MAPK8; O SQSTM1; P CAST; Q AIMP2; R RIPK1

Fig. 2

miRNAs associated with PRGs in LGGs. A Correlation heatmap of significantly associated PRGs and miRNAs, red represents positive correlation, blue represents negative correlation, and the shade of the color indicates the strength of the correlation; the darker the color, the stronger the correlation; B Network of miRNA and PRGs interactions, green circles represent PRGs, yellow diamonds represent miRNAs, and the thickness of the connecting lines indicates the strength of the correlation; C Sankey diagram of the regulatory relationship between miRNAs and PRGs, from PRGs to miRNAs to classification, the larger the width of the stream, the stronger the regulation

3.2 Identification of miRNA subtypes in LGG and differential expression of miRNAs in tumour subtypes

To further explore the possible indirect correlations involved in miRNAs in LGG, 91 miRNAs were clustered using NMF, and LGG was classified into two subtypes, C1 and C2 (Fig. 3A, B), and 85 differentially expressed miRNAs (DEmiRs) existed between the two subtypes as shown in Fig. 3C, D. The predominant diagnosis in the C1 group was "Oligodendroglioma, NOS", whereas the diagnosis in the C2 group was "Astrocytoma, anaplastic", and the proportion of patients receiving treatment was higher in the C2 group, with a significant difference in the proportion of patients receiving treatment between the two groups (Supplementary Table 3). Forty-five of these DEmiRs remained highly correlated with PRGs expression, and these miRNAs may be involved in a complex regulatory network in LGG and are closely linked to PRGs (Fig. 3E).

Fig. 3

Classification of LGG Subtypes Based on miRNAs Significantly Associated with PRGs. A NMF classification parameter map; B Consensus clustering map from NMF clustering, distinctly categorizing LGG patients into two subtypes; C miRNAs differentially expressed (DEmiRs) between the two subtypes, yellow and blue plots indicate up- and down-regulated miRs in one isoform relative to the other, respectively; D Heatmap of expression for these DEmiRs, red and blue colors represent C2 and C1 subtypes, respectively. E Venn diagram identifying 45 miRNAs that show differential expression between the two subtypes and are highly correlated with PRG expression

3.3 Construction of miRNA prediction model by lasso

To construct PRG-related miRNA prognostic features to predict the prognosis of LGG patients, univariate Cox regression analysis was first applied to screen miRNAs that were significantly associated with patients' overall survival (OS) (Fig. 4A). Subsequently, lasso further selected 15 prognostically relevant miRNAs. In total, 9 miRNAs and their corresponding coefficients were obtained (Fig. 4B, C), and the patient risk score was calculated using the following formula:

$$ } = - 0.00116618 \times } - } - 1296 + 0.001418971 \times } - } - 149 + 0.00548326 \times } - } - 155 + 0.000120955 \times } - } - 196} + 0.017413605 \times } - } - 222 + 0.006998872 \times } - } - 224 + 3.43489} - 05 \times } - } - 23} + - 0.112726538 \times } - } - 346 + 0.035499089 \times } - } - 616 $$

Fig. 4

Construction of miRNA prediction model for prognosis of LGG patients using LASSO regression analysis. A Forest plot to show the 15 prognostically relevant miRNAs obtained from one-way Cox analysis; B Ten-fold cross-validation to select the best log(lamda) in the TCGA dataset; C Trajectory plot of lasso coefficients; D K-M survival curves of the high and low risk groups in the TCGA dataset; E Time dependent ROC; F AUC of the TCGA dataset; G Risk factor plots of the TCGA dataset; H K-M survival curves of the high- and low-risk groups in the CGGA dataset; I Time-dependent ROC of the CGGA dataset; J AUC of the CGGA dataset; K Risk factor plots of the CGGA dataset

TCGA-miRNA expression profiles were used as the training set and CGGA-miRNA expression profiles as the validation set, and the optimal cut-off values were used to classify high-risk and low-risk groups. We evaluated the model with KM analysis, ROC curves, and risk factors to assess its prognostic potential in LGG patients. AUC areas greater than 0.6 at 1, 3, and 5 years were observed in both TCGA and CGGA for patients in the high-risk group (P < 0.05). It was found that the high-risk group had a significantly worse prognosis than the low-risk group, and the model expressed a more stable prognostic predictive ability (Fig. 4D–K).

3.4 LGG patients in different risk groups present different clinical features

In-depth analysis of the clinicopathological features of these risk groups revealed that patients in the high-risk group died more frequently, with different histological stages and IDH mutations (Fig. 5A, B), and that high-risk patients had histologically higher rates of mixed gliomas and Anaplastic Astrocytoma, and that the proportion of IDH mutations was higher in low-risk LGG patients (Fig. 5C, D). Were higher (Fig. 5C, D). These differences reveal the importance of Risk scores in the prognosis of LGG patients and provide new perspectives for future clinical interventions.

Fig. 5

Differences in Risk Score and Clinical Characteristics between patients of different risk levels. Heatmaps presenting the distribution of clinical characteristics between patients grouped by different risk levels in the (A) TCGA and (B) CGGA databases; grouped stacked bar charts presenting the frequency of (C) Histology and (D) IDH mutation statuses in patients grouped by different risk levels in these two datasets

3.5 Nomogram construction and validation

PMI scores were able to independently predict the prognosis of LGG patients both in univariate and multivariate Cox analyses (Fig. 6A, B). A Nomogram was constructed to predict the survival rate of LGG patients at 1,3,5 years (Fig. 6C), and the calibration curves (Fig. 6D) were used to assess the accurate predictive ability of the Nomogram for the prognosis, and the results of DCA indicated that the Nomogram was the best predictor of all the predictive factors optimally (Fig. 6E), a finding that suggests the potential of PMI-based construction of nomograms to provide valuable prognostic information for LGG patients.

Fig. 6

PMI-based construction of nomograms and evaluations. In LGG patients: A Univariate Cox analysis and (B) Multivariate Cox analysis; C construction of nomograms to predict prognosis; and (D) calibration curves

3.6 Differences in potential molecular mechanisms in LGG patients from different risk groups

After enrichment analysis of differential genes in LGG patients in high and low-risk groups, we found that these genes were significantly enriched in key biological processes such as extracellular matrix organisation, extracellular structure organisation, external encapsulating structure organisation and other key biological processes were significantly enriched (Fig. 7A). Emap plots (Fig. 7B) revealed high similarity between the significantly enriched functions (adj P < 0.001), which may be involved in LGG developmental processes through interactions. In addition, GSEA identified significantly enriched pathways (Fig. 7C), and we found that the Cytokine-cytokine receptor interaction and Ecm receptor interaction pathway was significantly enriched in the high-risk group (Fig. 7D, E), which suggests the inflammatory response and extracellular matrix remodelling in the high-risk patients' importance. On the contrary, the Ribosome pathway was significantly enriched in the low-risk group (Fig. 7F), suggesting that the role of protein synthesis processes in maintaining normal cellular function may be associated with a better prognosis.

Fig. 7

Functional enrichment analyses of high- and low-risk LGG groups. Differential genes between groups were analysed for (A) Gene ontology (GO) enrichment; B Emap plot showing pairwise similarity between significantly enriched terms; C Mountain range plot demonstrating significantly enriched KEGG pathways; D Cytokine-cytokine receptor interaction; E Ecm receptor interaction; F GSEA map of Ribosome. BP biological process, CC cellular component, MF molecular function

3.7 Association of immune microenvironment heterogeneity with low-grade glioma risk score

The proportion of different types of immune cells was significantly different between the high-risk and low-risk LGG groups (Fig. 8A). It was especially noticeable that the infiltration score for the high-risk group was significantly higher than for the low-risk group, while the response score for the high-risk group was significantly lower. In addition, correlation analysis with Riskscore revealed that immune cells, including CD4 native, iTreg, NK, and MAIT cells, showed a significant correlation with the risk score, whereas the Response score showed a significant inverse correlation (Fig. 8B). The proportion of immune response was even lower in the high-risk group of patients (Fig. 8C). This finding suggests that the tumour microenvironment of high-risk LGG patients may have more immunosuppressive properties, which may have contributed to the low responsiveness to immunotherapy.

Fig. 8

Correlation between immune microenvironment heterogeneity and risk scores for low-grade gliomas. A The proportions of 18 T and B cells, NK cells, monocyte cells, macrophage cells, neutrophil cells, and DC cells, and the corresponding immune scores; B Scatterplot of correlation of Riskscore with 24 immune cells; C Comparison of proportions of immune response in high and low risk groups

3.8 Potential drug candidates for LGG treatment

Drug sensitivity analyses showed significant differences in the name-sensitivity of some common drugs, such as Axitinib, Bortexomib, and Vorinostat, among others, in patients with different risk classes (Fig. 9A, B). Further CMap analyses showed that different compounds presented variable strengths of action in common tumour cell lines (Fig. 9C), and as shown in Fig. 9D, fasudil obtained the lowest combination score, implying that it possesses the potential to reverse the disease state of high-risk LGG patients, and is thus a promising candidate for the treatment of patients with poor prognosis LGG (see Fig. 9D). These findings are valuable in guiding personalised treatment regimens for LGG.

Fig. 9

Comparison of drug sensitivity and drug prediction. Drug sensitivity of common drugs for LGG patients in high and low risk groups was assessed using (A) pRRophetic and (B) oncoPredicti; C relationship between molecularly targeted Therapy and common cell lines was assessed using CMap; and D predicted top 5 most likely therapeutic drugs

3.9 Expression levels of 9 miRNAs and their prognostic profiles

We found that, as shown in Fig. 10, except for hsa-miR-1296 and hsa-miR-346, other miRNAs (hsa-miR-149, hsa-miR-155, hsa-miR-196b, hsa-miR-222, hsa-miR-224, hsa-miR-23a, hsa-miR- 616) were expressed at significantly higher levels in high-risk patients than in low-risk patients. Moreover, except for these two miRNAs, the other seven miRNAs could be pro-oncogenic, and patients with high expression levels had shorter survival, which could be used as a marker of poor prognosis in LGG patients (Fig. 11), while hsa-miR-1296 and hsa-miR-346 could be used as cancer-suppressing miRNAs, which might be important for maintaining normal cellular function and preventing malignant progression in LGG significance. Using the primer sequences in Table 2 for qPCR and finally normalised by U6, we found the same miRNA expression results in the cells (Fig. 12).

Fig. 10

Expression levels of 9 miRNAs. patients in high and low risk groups (A) hsa-miR-149; B hsa-miR-155; C hsa-miR-196b; D hsa-miR-222; E hsa-miR-224; F hsa-miR-23a; G hsa-miR-1296; H hsa-miR-346; I hsa- miR-616’s RNA-seq expression levels of miR-616

Fig. 11

K-M survival curves of the 9 miRNAs. A hsa-miR-149; B hsa-miR-155; C hsa-miR-196b; D hsa-miR-222; E hsa-miR-1296; F hsa-miR-224; G hsa-miR-23a; H hsa-miR-346; I hsa-miR-616 high expression and low expression K-M survival curves of LGG patients

Table 2 Primer sequences for miRNAFig. 12

Expression of 9 miRNAs between Human neural astrocytes and glioma cells. Relative expression levels of miR-1296, miR-346, miR-149, miR-155, miR-222, miR-224, miR-23a, miR-616 in human neural astrocytes and T98G, HS 683 were detected using qPCR