Integrative analysis of tissue and circulating miRNAs as biomarkers for progression and survival in hepatocellular carcinoma.

1
Introduction
HCC is the most common type of primary liver cancer, accounting for over 90 % of all primary liver tumor cases. It primarily develops in the setting of chronic liver disease and affects approximately 80–90 % of individuals with cirrhosis. Globally, HCC ranks as the fifth most prevalent cancer and the second leading cause of cancer-related death in men, following lung cancer [1]. Despite improvements in detection and management, the prognosis for HCC remains poor, with a five-year survival rate of just 18 %, second only to pancreatic cancer in terms of lethality. Major risk factors include chronic viral hepatitis (hepatitis B and C), alcoholic liver disease, and non-alcoholic steatohepatitis (NASH) [1]. Globally, HCC is the fifth most common disease and the second greatest cause of cancer-related death in men, after lung cancer. Despite advances in identification and treatment, the prognosis for HCC remains low, with a five-year survival rate of about 18 %, ranking second only to pancreatic cancer in terms of lethality. Chronic viral hepatitis (hepatitis B and C), alcoholic liver disease, and non-alcoholic steatohepatitis (NASH) are all significant risk factors [1].
HCC is frequently diagnosed at an advanced stage because it develops without apparent signs in the early stages. As a result, many patients are no longer eligible for potentially curative therapies such as surgical resection or liver transplantation when they are diagnosed. The disease's aggressive behavior—characterized by rapid development, early vascular invasion, and underlying liver damage—makes treatment particularly difficult and adds to its poor overall prognosis [2]. In a study conducted between 2004 and 2008, HCC had a 5-year survival rate of 23.3 %, indicating an extremely negative prognosis when compared to other cancers [3].
MicroRNAs (miRNAs) are small non-coding RNA molecules that play a role in post-transcriptional regulation of gene expression. In cancer, including hepatocellular carcinoma, miRNA expression is often altered through mechanisms such as gene amplification or deletion, epigenetic changes or defects in the miRNA processing pathway. These changes can affect key cellular functions [4]. As described in Hanahan and Weinberg's framework of cancer hallmarks—such as sustained proliferative signaling, evasion of growth suppressors, resistance to cell death, and increased invasion and metastasis. Dysregulated miRNAs have been shown to impact many of these pathways by acting as oncogenes or tumor suppressors depending on the genes they target [5]. In HCC, miRNAs have been implicated in multiple aspects of tumor development and progression. Some are upregulated and contribute to tumor growth and metastasis, while others are downregulated and normally suppress these processes. Because miRNAs often reflect both the origin and behavior of tumor cells, they are being explored as potential biomarkers for diagnosis, prognosis and even as therapeutic targets [6]. Although many individual miRNAs have been studied in HCC, comprehensive analyses that capture the full landscape of miRNA expression using large-scale public datasets are limited. Much of the existing literature is based on studies with small cohorts, varied analytical approaches or limited clinical information which makes it difficult to draw consistent conclusions or apply findings in a clinical setting [[7], [8], [9]]. Moreover, the functional significance and regulatory networks of many dysregulated miRNAs in HCC are not well understood [10]. Beyond tumor tissue, circulating serum miRNAs have become non-invasive biomarkers for cancer. These are very stable in blood and change dynamically with disease progression so are useful for early detection, prognosis and monitoring treatment response [11]. Although there are not many large-scale studies comparing serum miRNA profiles across different stages of liver disease – from chronic hepatitis and cirrhosis to HCC, a recent study by Jeng et al. revealed that specific circulating miRNAs can independently predict outcomes in HCC [12]. To address these gaps, we performed a systematic bioinformatics analysis of liver cancer data from The Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA-LIHC) to identify differentially expressed (DE) miRNAs between tumor and healthy liver tissues. Additionally, we integrated these findings with target gene prediction, functional enrichment analyses, miRNA–gene interaction network construction and clinical correlation studies including survival and tumor stage analyses. In this study, we analyzed serum miRNA data from GSE113740 and tumor derived miRNA data from TCGA-LIHC so we can have a complete picture. Overall, our goal was to unravel biologically and clinically relevant miRNA signatures that can potentially serve as prognostic biomarkers and therapeutic HCC targets.

2
Materials and methods
This study utilized two datasets: miRNA sample data of liver hepatocellular carcinoma TCGA-LIHC and serum miRNA dataset (GSE113740) and the analysis was performed in two phases.
2.1
TCGA liver hepatocellular carcinoma (LIHC) dataset
2.1.1
Data selection and preprocessing
For the first phase, miRNA sample data of liver hepatocellular carcinoma LIHC were obtained from The Cancer Genome Atlas (TCGA Research Network): https://www.cancer.gov/tcga. The dataset included a total of 425 samples, consisting of 376 tumor tissues and 49 matched healthy liver tissues.
All analyses were performed using R (v4.5.0) and Python (v3.11) with key packages including limma (v3.50.0) for differential expression analysis, survival (v3.3-1) for Kaplan-Meier survival analysis, clusterProfiler (v4.2.0) for GO/KEGG enrichment, igraph (v1.2.6) and networkx (v2.6.3) for miRNA-gene network construction, ggplot2 (v3.3.6) and matplotlib (v3.4.3) for visualization, pheatmap (v1.0.12) for heatmaps, and pandas (v1.3.0) and numpy (v1.21.0) for data manipulation.

2.1.2
Sequence reads processing and differential expression analysis
The Galaxy platform was utilized to conduct a meta-analysis comparing two groups, HCC and normal samples, using its intuitive and user-friendly interface. Galaxy is an open-source bioinformatics tool that facilitates reproducible and accessible analysis. The preprocessing steps included data normalization, log2 transformation, and removal of miRNAs, ensuring high-quality data input for subsequent analyses. DE miRNAs between tumor and normal tissues were identified using the limma-trend (v3.50.0) statistical method implemented within Galaxy through R. This method was particularly appropriate given the log-transformed nature of the expression data, allowing robust and reliable modeling of differential expression.
Differentially expressed miRNAs were identified using an adjusted p-value threshold of ≤0.05 and |log2 fold change| > 1. The p-value threshold controls for false discovery rate, while the log2FC threshold ensures selection of miRNAs with biologically meaningful expression differences between tumor and control samples.

2.1.3
Target prediction
The miRNA target prediction was performed for the top 10 DE miRNAs based on their statistical significance (adjusted p-values). miRTarBase, a curated database of experimentally validated miRNA–target interactions were utilized to retrieve corresponding gene targets. Pandas (v1.3.0) DataFrame, and miRNA identifiers were standardized to match the format used in miRTarBase. Validated gene targets were retrieved for each miRNA from miRTarBase and used for downstream analysis. The networkx library in python (v2.6.3) was deployed for the construction of overall interaction network and simplified network construction focused on high confidence, frequently occurring target genes for clearer interpretation.

2.1.4
Functional enrichment analysis
Functional enrichment analysis was conducted using the DAVID (Database for Annotation, Visualization, and Integrated Discovery) tool. Predicted and validated gene targets derived from miRNA interactions were submitted to DAVID for analysis. GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Gene and Genomes) pathway enrichment results were obtained. Enrichment plots emphasized several critical biological processes, including signal transduction, apoptosis, and regulation of the cell cycle, pathways frequently associated with cancer initiation and progression.

2.1.5
Survival analysis
The survival analysis was performed to explore the prognostic relevance of top 10 DE miRNAs using Kaplan-Meier analysis. Patients were stratified into two groups. High expression and low expression based on median expression values of each miRNA and survival differences were plotted using log-rank tests. Plots were generated using the lifelines (v0.25.6) Python package.

2.1.6
Clinical subgroup analysis
To identify miRNAs associated with disease progression, we conducted differential expression analysis between clinical subgroups. In the T-stage based grouping patients were divided into early T1/T2 and advanced stage T3/T4 groups and in Metastasis-based grouping the patients were divided into M0/M1 stage. Using the same limma-trend results the DE miRNAs were identified and visualized using bar plots and box plots.

2.2
Serum miRNA Dataset Analysis (GSE113740)
2.2.1
Data selection and preprocessing
In the second phase of this study, circulating serum miRNA expression data were obtained from the Gene Expression Omnibus (GEO) database under the accession number GSE113740 [13]. The dataset included miRNA expression profiles from four clinically distinct groups: 969 normal controls, 46 chronic hepatitis patients, 93 liver cirrhosis patients, and 345 HCC patients. The data files included the normalized and log2-transformed expression matrix. These datasets were processed and analyzed using R (v4.5.0) and Python (v3.11), leveraging a comprehensive suite of packages. In R, key packages included limma (v3.50.0) for differential expression analysis, survival (v3.3-1) for Kaplan-Meier survival analysis, clusterProfiler (v4.2.0) for GO and KEGG enrichment, igraph (v1.2.6) for network construction, ggplot2 (v3.3.6) and pheatmap (v1.0.12) for data visualization, and dplyr (v1.0.8) and tidyr (v1.1.4) for data manipulation. In Python, tools such as networkx (v2.6.3), matplotlib (v3.4.3), seaborn (v0.11.2), lifelines (v0.25.6), pandas (v1.3.0), and numpy (v1.21.0) were used for visualization, network analysis, and survival modeling.

2.2.2
Sequence Reads processing and differential expression analysis
The differential expression analysis of the serum miRNA dataset was conducted using the limma-trend approach, executed in R with the limma package (v3.64.0).
Samples were categorized into Control, Chronic Hepatitis, Cirrhosis, and HCC groups (NControl = 969, NChronic Hepatitis = 46 NCirrhosis = 93 and NHCC = 345) based on clinical metadata. Pairwise comparisons were conducted between HCC vs Control, Chronic Hepatitis vs Control, and Cirrhosis vs Control. MiRNAs were classified as DE if they exhibited an adjusted p-value (FDR) < 0.05. Additionally, fold change (FC) values were used to separate miRNAs into overexpressed (log2FC > 1) or underexpressed (log2FC < 1) groups. A total of 2238 DE miRNAs were identified in the HCC vs Control comparison, 1350 in Chronic Hepatitis vs Control, and 1282 in Cirrhosis vs Control. A predominance of overexpressed miRNAs was observed across all disease stages.

2.2.3
Visualization and heatmap construction
To visually illustrate the differential expression profiles, volcano plot was generated for each disease comparison. This plot displayed the magnitude and significance of miRNA expression changes. Additionally, a combined heatmap was created featuring the top 50 most significantly DE miRNAs across the four groups (Control, Chronic Hepatitis, Cirrhosis, and HCC). The heatmap used a Red-Blue diverging color palette to distinguish between upregulated and downregulated miRNAs, providing a clear visualization of expression patterns associated with liver disease progression.

3
Results
3.1
Differential expression analysis of TCGA liver hepatocellular carcinoma (LIHC) dataset
We analyzed 1882 miRNAs from HCC tissues and applied a strict cutoff (adjusted p ≤ 0.05 and |log2FC| > 1) to identify those that were truly dysregulated. Using this threshold, we found a focused set of 16 miRNAs that showed significant expression changes between tumour and non-tumour tissues (Fig. 1).
Among them, hsa-miR-1258 (log2FC = −2.12, adj. p = 3.6 × 10−83), hsa-miR-490 (−2.16, adj. p = 3.0 × 10−76), and hsa-miR-383 (−1.53, adj. p = 1.4 × 10−31) were the most strongly downregulated. Other miRNAs like hsa-miR-4686, hsa-miR-6718, and hsa-miR-1247 also showed consistent decreases.
On the other hand, several miRNAs were upregulated in tumors. For example, hsa-miR-767 (+1.33, adj. p = 3.4 × 10−27), hsa-miR-105-1/2, hsa-miR-34c, and hsa-miR-891a were all significantly higher in HCC tissues, suggesting potential oncogenic roles.
From this set, we selected the top 10 most strongly altered miRNAs (with the largest fold-changes and strongest statistical support) for further pathway and survival analysis (Table 1). These represent the most promising candidates likely driving HCC progression (see Table 2).

3.2
Functional enrichment of target genes
To understand what the top 10 differentially expressed miRNAs might be doing in HCC, we looked at the biological processes and pathways their target genes are involved in.
GO analysis showed that these target genes are linked to neuron growth and development, synapse formation, and organ development. The top processes included axon formation, regulation of neuron projections, gland development, forebrain development, synapse structure and organization, dendrite development, and respiratory system development. This suggests that these miRNAs could influence not just cancer cell behavior but also processes related to tissue and cellular organization (Fig. 2).
KEGG pathway analysis highlighted key signaling pathways that are often active in cancer. The top 10 pathways were PI3K-Akt signaling, HPV infection, MAPK signaling, Endocytosis, Salmonella infection, cAMP signaling, Focal adhesion, Neurotrophin signaling, Glutamatergic synapse, and Prostate cancer. Many of these, such as PI3K-Akt and MAPK, are important for cell survival, growth, and movement, implying that these miRNAs may help drive tumor progression by affecting these pathways (Fig. 3).
Overall, these results show that the top DE miRNAs in HCC are likely regulating important processes for cell growth, signaling, and tissue organization, which could be key for understanding their role in liver cancer. The top-ranked GO terms were derived from the combined targets of the top 10 differentially expressed miRNAs. As shown in (Fig. 2).
Target prediction analysis identified several high-confidence gene–miRNA interactions (score >0.95). For instance, hsa-miR-1258 was predicted to regulate multiple genes involved in transcriptional control and RNA processing, including ZNF571, PTBP3, EBF2, and SEC23B. Many of these targets have recognized roles in oncogenic pathways such as cell-cycle regulation, protein trafficking, and epithelial–mesenchymal transition, supporting the putative involvement of miR-1258 in hepatocarcinogenesis. These findings provide mechanistic insight into the downstream effects of dysregulated miRNAs identified in this study. Comprehensive gene-miRNA interaction for the top highly expressed miRNA is given in (Supplementary File-1).

3.3
Survival analysis
Kaplan–Meier survival analysis was performed to assess the prognostic value of the top ten HCC-related DE miRNAs (hsa-miR-1258, hsa-miR-490, hsa-miR-383, hsa-miR-767, hsa-miR-105-2, hsa-miR-105-1, hsa-miR-6718, hsa-miR-5589, hsa-miR-1247, and hsa-miR-187). Among these, only hsa-miR-187 showed a statistically significant association with overall survival (p = 0.0324) (Fig. 4). Higher expression of hsa-miR-187 was linked to poorer survival outcomes, suggesting a potential role as a risk-associated biomarker in HCC. Moreover hsa-miR-33b (p = 0.0016) and hsa-miR-424 (p = 0.029) showed significance in the survival. The remaining miRNAs did not show significant correlations with patient survival in this cohort.
3.3.1
Cox regression analysis and hazard ratios of candidate miRNAs
To further evaluate the prognostic significance of the top 10 HCC-related miRNAs, we performed univariate and multivariate Cox proportional hazards regression analyses.
In the univariate analysis, hsa-miR-6718 (HR = 1.34, 95 % CI: 1.10–1.63, p = 0.0037) and hsa-miR-187 (HR = 1.25, 95 % CI: 1.08–1.45, p = 0.0031) were significantly associated with poorer overall survival. Other miRNAs showed no statistically significant associations in this cohort.
Multivariate analysis adjusting for relevant clinical covariates confirmed the independent prognostic value of several miRNAs. hsa-miR-6718 (HR = 1.29, 95 % CI: 1.01–1.66, p = 0.045) and hsa-miR-187 (HR = 1.40, 95 % CI: 1.18–1.66, p < 0.001) remained significantly associated with increased risk, while hsa-miR-5589 showed a protective effect (HR = 0.88, 95 % CI: 0.77–0.99, p = 0.038). The other miRNAs did not reach statistical significance in the multivariate model. Neither hsa-miR-424 (HR = 1.11, 95 % CI: 0.92–1.33, p = 0.28) nor hsa-miR-33b (HR = 1.18, 95 % CI: 0.98–1.44, p = 0.088) was significantly associated with overall survival.
When combining Kaplan–Meier and Cox regression analyses, we observed that some miRNAs, such as hsa-miR-6718 and hsa-miR-5589, showed significance in the Cox model but not in Kaplan–Meier analysis. This difference arises because Kaplan–Meier analysis relies on dichotomizing patients into high- and low-expression groups, which may overlook more subtle associations. In contrast, the Cox proportional hazards model evaluates expression as a continuous variable and accounts for covariates, offering greater sensitivity in detecting prognostic signals. Taken together, the complementary use of these methods underscores hsa-miR-187 and hsa-miR-6718 as potential risk-associated biomarkers and highlights hsa-miR-5589 as a candidate with a protective role in HCC.

3.4
Clinical subgroup analysis
To explore the clinical relevance of miRNA expression further, we also performed differential expression analysis. Among primary clinical subgroups, i.e., metastasis status and T-stage of tumors. With limma-trend analysis, we detected miRNAs correlated with disease progression.
3.4.1
T-stage classification (T1/T2 vs. T3/T4)
Our research identified hsa-miR-106b to be greatly upregulated in late-stage tumors (T3/T4). relative to early-stage tumors (T1/T2). With a p-value < 0.05. The overexpression of hsa-miR-106b in aggressive tumors indicates its potential role in tumor progression. Thereby suggesting that it is a stage-specific marker in HCC (Fig. 5).

3.4.2
Metastasis-based grouping (M0 vs. M1)
Some microRNAs exhibited differential expressions between metastatic (M1) and non-metastatic (M0) samples. Interestingly, hsa-miR-124-1, hsa-miR-124-2, hsa-miR-124-3, hsa-miR-3184, and hsa-miR-4641 were the most dysregulated microRNAs. These findings are shown in Fig. 6.
These findings contribute to further evidence for the involvement of miRNA in disease aggressiveness and for their potential application in the stratification of patients. According to clinical risk factors. Patients were classified as having (M1) or not having (M0) distant metastasis. Differential analysis revealed a cluster of miRNAs with strong correlation to metastatic status, suggesting their potential functions in HCC occurrence.

3.5
Serum miRNA Dataset Analysis
To complement the tissue analysis, we examined circulating miRNAs from the GSE113740 dataset (HCC, non-cancer, cirrhosis, and hepatitis). Several serum miRNAs were significantly dysregulated (adj. p ≤ 0.05).
In HCC vs non-cancer, upregulated miRNAs such as hsa-miR-5100, hsa-miR-195-5p, and hsa-miR-1290 were enriched for targets involved in cell cycle, endocytosis, and proteoglycans in cancer, with GO terms linked to chromosome segregation and nucleocytoplasmic transport (Fig. 7). In HCC vs cirrhosis, hsa-miR-422a and hsa-miR-204-3p were upregulated, while hsa-miR-4687-3p and hsa-miR-3619-3p were downregulated. Pathways included cell cycle, ER protein processing, and senescence, with GO enrichment in Wnt signaling and nuclear transport (Fig. 8). In HCC vs hepatitis, upregulated miRNAs such as hsa-miR-1185-2-3p and hsa-miR-6132 contrasted with downregulated hsa-miR-6778-5p and hsa-miR-3619-3p. Enriched pathways involved viral carcinogenesis, adherens junction, and cell cycle, alongside GO terms for RNA localization, nuclear transport, and autophagy.
Overall, serum-derived miRNAs converged on pathways also highlighted in tissue analysis (e.g., cell cycle, nucleocytoplasmic transport, Wnt signaling), reinforcing their promise as non-invasive biomarkers for HCC progression.
3.5.1
Visualization of (DE) miRNAs
A combined heatmap of the top 50 DE miRNAs effectively distinguished Normal samples from Chronic Hepatitis, Cirrhosis, and HCC groups based on their unique expression patterns. Overall, the distinct serum miRNA profiles identified highlight the potential role of circulating miRNAs as non-invasive biomarkers for early detection and monitoring of liver diseases (Fig. 9).
A volcano plot was also generated to visualize the DE miRNAs across HCC, Liver Cirrhosis, and Chronic Hepatitis compared to Normal controls. Each disease group demonstrated a distinct distribution of miRNAs based on log2FC and adj p-value significance. Overall, most miRNAs were significantly overexpressed across all the groups, particularly in HCC samples, indicating strong miRNA dysregulation associated with the disease progression (Fig. 10).
The volcano plot clearly demonstrated that many miRNAs were upregulated (+ve FC) while a smaller proportion were under expressed (-ve FC) in diseased conditions relative to control samples.

3.5.2
Association of circulating miRNAs with AFP levels and cancer stage
To further evaluate the clinical relevance of the dysregulated serum miRNAs in HCC, we examined their correlation with serum alpha-fetoprotein (AFP), a key biomarker for liver cancer, and their association with cancer stage and Child-Pugh class.
Among the nine serum miRNAs analyzed, hsa-miR-3619-3p showed the strongest negative correlation with AFP (rho = −0.28, p = 1.07 × 10^-9), followed by hsa-miR-1290 (rho = −0.26, p = 4.79 × 10^-8) and hsa-miR-4687-3p (rho = −0.21, p = 1.06 × 10^-5). In contrast, hsa-miR-1185-2-3p (rho = 0.27, p = 1.12 × 10^-8), hsa-miR-422a (rho = 0.25, p = 8.71 × 10^-8), and hsa-miR-204-3p (rho = 0.17, p = 3.74 × 10^-4) displayed positive correlations with AFP levels (Table 3) (Fig. 11). The remaining miRNAs, including hsa-miR-5100, hsa-miR-195-5p, and hsa-miR-6132, did not show statistically significant associations with AFP (p > 0.1). These results indicate that a subset of serum miRNAs may reflect tumor burden in HCC patients, providing complementary information to AFP measurements.
In addition, we evaluated miRNA expression across pathological cancer stages using Kruskal-Wallis tests. Several miRNAs exhibited stage-dependent expression patterns, most notably hsa-miR-5100, hsa-miR-195-5p, hsa-miR-1290, hsa-miR-422a, hsa-miR-1185-2-3p, and hsa-miR-3619-3p (adjusted p-values <1 × 10^-40). Similarly, hsa-miR-204-3p, hsa-miR-4687-3p, and hsa-miR-422a displayed significant differences across Child-Pugh classes (adjusted p < 5 × 10^-9), highlighting their potential relevance to liver function and disease severity (Table 3) (Fig. 11).
Collectively, these findings reveal that several circulating miRNAs not only recapitulate dysregulation observed in HCC tissue but also correlate with established clinical parameters, including AFP levels and cancer stage. This dual relationship, reflecting both molecular and clinical features, underscores the novelty of our integrative approach, distinguishing it from prior studies that primarily focused on differential expression without linking serum miRNAs to key biomarkers or disease progression. Importantly, miRNAs such as hsa-miR-3619-3p, hsa-miR-1290, and hsa-miR-1185-2-3p emerge as particularly promising candidates for non-invasive monitoring of HCC progression.

4
Discussion
4.1
Tissue miRNAs and prognostic relevance
Hepatocellular carcinoma (HCC) remains one of the most lethal cancers worldwide, and the lack of reliable biomarkers for early detection and prognosis contributes to its poor clinical outcomes [14,15]. In this study, we conducted an integrative bioinformatics analysis of tissue and circulating miRNA expression profiles to identify miRNAs associated with tumor progression, metastasis, and clinical outcomes. Using stringent statistical thresholds and multivariate survival models, we identified 16 significantly dysregulated miRNAs in HCC tissues, of which hsa-miR-187, hsa-miR-6718, and hsa-miR-5589 exhibited strong prognostic relevance. Specifically, hsa-miR-187 and hsa-miR-6718 were independently associated with poorer overall survival, while hsa-miR-5589 showed a protective effect. These findings refine and expand upon previous in silico-only studies that often reported larger but less specific candidate lists [10,16].
Our results highlight the biological and clinical significance of these miRNAs. In our study overexpression of hsa-miR-187 has been shown to be associated with poor prognosis in HCC patients this is in contrast with previous study by Han et al. which showed that miR-187 as a tumor suppressor by targeting IGF-1R pathway [17]. However, it has been shown previously that overexpression of miR-187 is associated with poor prognosis in several cancers consistent with our study [18]. The novel association of hsa-miR-6718 with poor survival suggests a potential regulatory role in tumor cell proliferation or metastasis, although its function in hepatic tissue has not yet been experimentally characterized. Conversely, hsa-miR-5589 demonstrated a consistent protective association, potentially via modulation of PI3K–Akt and MAPK signaling cascades, both of which are critical to hepatocarcinogenesis [[19], [20], [21]]. These mechanistic insights align with our functional enrichment analyses, where target genes of the top miRNAs were significantly enriched in cancer-related pathways such as PI3K–Akt, MAPK, focal adhesion, and nucleocytoplasmic transport—processes central to tumor proliferation, migration, and survival.
Clinical subgroup analyses further revealed that hsa-miR-106b was upregulated in advanced-stage tumors (T3/T4), supporting its association with aggressive disease phenotypes. Previous studies have shown that miR-106b directly targets and inhibits RUNX3 a transcription factor and thereby promoting oncogenesis. [22]. Similarly, members of the hsa-miR-124 family were found to be downregulated in metastatic tumors, consistent with their reported tumor-suppressive effects through inhibition of CRKL and ERK signaling in HCC [23]. Together, these results underscore the relevance of specific miRNA signatures for patient stratification by disease stage and metastatic potential. Importantly, by incorporating hazard ratios and multivariate Cox modeling, our analyses provide a more robust and clinically relevant assessment of prognostic significance compared with previous studies relying solely on univariate survival testing.

4.2
Circulating miRNAs and clinical translation
The integration of serum miRNA data from GSE113740 provides additional translational insight, emphasizing the potential of circulating miRNAs as minimally invasive biomarkers for disease monitoring. Several serum miRNAs—most notably hsa-miR-3619-3p, hsa-miR-1290, and hsa-miR-1185-2-3p—were differentially expressed between HCC and control groups and correlated significantly with clinical parameters such as alpha-fetoprotein (AFP) levels, tumor stage, and Child–Pugh classification. hsa-miR-3619-3p and hsa-miR-1290 exhibited strong negative correlations with AFP, suggesting that their decreased serum expression may reflect increasing tumor burden. Conversely, hsa-miR-1185-2-3p and hsa-miR-422a were positively associated with AFP and advanced stages, implying a potential role in tumor progression.
The clinical potential of this is backed up by previous work. Wang et al. showed a serum miRNA signature including miR-486-5p was an independent prognostic factor for recurrence, and its combination with AFP dramatically improved predictive accuracy (AUC 88.02 %) [24]. miRNA panels are diagnostic; Umezu et al. showed circulating miRNA patterns could distinguish HCC from biliary tract cancer (AUC 0.754) and even identify the etiology of HCC (e.g. post-viral cure) [25]. This evidence strongly supports the development of the multi-miRNA panels similar to our work shown. Prior studies have demonstrated that circulating miRNAs can serve as dynamic indicators of tumor activity due to their stability and detectability in blood [12,26]. Our results strengthen this concept by showing that specific serum miRNAs mirror tissue-level dysregulation and disease severity. Importantly, several identified serum miRNAs—particularly miR-3619-3p—have limited prior reporting in HCC, whereas miR-1290 has been reported in HCC (including serum/exosomal contexts), though it has not been broadly evaluated alongside AFP and clinical staging in an integrated tissueand serum framework as presented here [[11], [27], [28], [29]].
However, it is crucial to interpret these results as hypothesis-generating. While the consistent correlation between circulating miRNA levels, AFP, and clinical staging suggests biological relevance, experimental and longitudinal validation is essential before clinical implementation. The observed overlap between tissue and serum signatures supports the biological plausibility of using circulating miRNAs as surrogates for tumor-derived dysregulation, but their diagnostic specificity and predictive performance must be evaluated in larger, independent cohorts.
To put our findings into the context of HCC miRNA research, Table 4 provides a comprehensive comparison that reveals both concordant and novel insights. While our results for well-known miRNAs like miR-21, miR-122 and the miR-124 family match the literature, we also found some discrepancies and new associations. Notably, we distinctly characterized hsa-miR-187 as an independent risk factor whereas some previous studies presented it as a tumor suppressor, highlighting the context dependent nature of miRNA function. We also found several under-investigated serum miRNAs, including hsa-miR-3619-3p, hsa-miR-1185-2-3p, and hsa-miR-4687-3p, all of which are strongly correlated with AFP and disease stage, underscoring their potential as novel non-invasive biomarkers. This comparison validates our approach and pinpoints the miRNAs that need to be further investigated.

4.3
Limitations, comparison with prior studies, and future directions
This study's integrative design offers distinct advantages over earlier in silico analyses that focused exclusively on either tissue or circulating data. By combining TCGA tissue profiles with GEO serum datasets, we identified both tumor-intrinsic and systemic miRNA alterations, providing a more comprehensive view of HCC biology. Previous studies (e.g., Huang et al., 2021; Feng et al., 2023) examined either tumor-derived or serum miRNAs in isolation, whereas our work reveals convergence between both compartments—particularly within pathways governing cell cycle regulation, nucleocytoplasmic transport, and MAPK signaling. This cross-validation across biological sources enhances the robustness and translational potential of the findings.
Nonetheless, several limitations must be acknowledged. First, this analysis is based entirely on publicly available datasets, which may vary in sample processing, patient demographics, and sequencing platforms, potentially introducing batch or selection bias. Second, our conclusions rely on bioinformatics-based target prediction and enrichment tools; experimental validation is necessary to confirm causal mechanisms. Third, survival data were unavailable for serum samples, restricting direct prognostic assessment of circulating miRNAs. Finally, although we applied multivariate models to adjust for clinical covariates, the lack of comprehensive clinical metadata in TCGA may still limit interpretation.
Future research should focus on functional validation of key miRNAs—particularly hsa-miR-187, hsa-miR-6718, and hsa-miR-5589 in patient samples and cell lines based models to elucidate their mechanistic roles in hepatocarcinogenesis. Large-scale, prospective studies integrating transcriptomic, proteomic, and epigenetic data could further delineate the molecular networks underlying HCC progression. Ultimately, this integrative bioinformatics framework lays the groundwork for developing multi-marker miRNA panels that may improve early diagnosis, prognosis, and therapeutic monitoring in hepatocellular carcinoma.

5
Conclusion
This study highlights a focused group of miRNAs with important roles in hepatocellular carcinoma. From tissue analysis, hsa-miR-187 and hsa-miR-6718 were associated with poor survival, while hsa-miR-5589 showed a protective effect, supporting their potential as prognostic markers. Clinical subgroup analyses further identified hsa-miR-205 as a marker of advanced stage and members of the hsa-miR-124 family as indicators of metastatic potential.
Circulating miRNAs provided complementary insights. Several, including hsa-miR-3619-3p, hsa-miR-1290, and hsa-miR-1185-2-3p, correlated with AFP levels and disease stage, suggesting their value as non-invasive biomarkers for monitoring tumor burden and disease progression.
By integrating tissue and serum datasets, this work offers a clearer and more clinically relevant picture of miRNA dysregulation in HCC. The findings support the development of miRNA-based biomarkers for early detection, prognosis, and patient stratification, though validation in larger, prospective cohorts remains necessary.

CRediT authorship contribution statement
Abdullah Jabri: Writing – review & editing, Writing – original draft, Formal analysis, Data curation, Conceptualization. Abdulaziz Mhannayeh: Writing – review & editing, Writing – original draft, Visualization, Formal analysis, Data curation. Mohamed Alsharif: Writing – review & editing, Writing – original draft, Formal analysis, Data curation. Bader Taftafa: Writing – review & editing, Writing – original draft, Visualization, Methodology, Data curation. Tooba Mujtaba: Writing – review & editing, Writing – original draft, Visualization, Software, Methodology, Formal analysis, Data curation. Arshiya Akbar: Writing – review & editing, Writing – original draft, Visualization, Software, Methodology, Formal analysis, Data curation. Ahmed Abu-zaid: Writing – review & editing, Writing – original draft, Visualization, Formal analysis, Conceptualization. Ahmad Mir Tanveer: Writing – review & editing, Writing – original draft, Visualization, Methodology, Formal analysis. Mohammad Imran Khan: Writing – review & editing, Writing – original draft, Visualization, Supervision, Data curation, Conceptualization. Firoz Ahmed: Writing – review & editing, Writing – original draft, Validation, Supervision, Methodology, Formal analysis, Conceptualization. Itika Arora: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Formal analysis. Ahmed Yaqinuddin: Writing – review & editing, Writing – original draft, Validation, Supervision, Methodology, Formal analysis, Data curation, Conceptualization.

Ethics approval and consent to participate
This study used publicly available, de-identified data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. Ethical approval and patient consent were not required.

Availability of data and materials
All data analyzed in this study are publicly available.•TCGA-LIHC dataset: https://www.cancer.gov/tcga

•GEO accession GSE113740: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE113740

Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.