The present study used a bioinformatics-guided workflow to identify candidate hub genes in breast cancer and subsequently evaluated selected genes and literature-supported miRNAs in a clinical tissue cohort. CDC20 and MELK emerged as the only hub genes shared by the Grade II and Grade III PPI modules. Their increased expression in breast cancer was supported by external GOBO analysis and by the direction of the qRT-PCR findings in the validation cohort. However, the clinical data did not demonstrate statistically significant differences in CDC20 or MELK expression between Grade II and Grade III tumors. This distinction is important because the bioinformatics analysis identifies candidates associated with tumor-related transcriptional changes, whereas the relatively small experimental cohort did not provide statistical evidence of a grade-specific difference.
CDC20 is a co-activator of the anaphase-promoting complex/cyclosome (APC/C) and contributes to mitotic progression and chromosome segregation (
11,
12). Dysregulation of CDC20 has been implicated in chromosomal instability, abnormal cell-cycle progression, and malignant behavior (
13,
14). The increased tumor-associated expression observed in the present study is therefore biologically consistent with previous literature. Nevertheless, because the Grade II-versus-Grade III comparison was not significant (P = 0.820), our data do not establish CDC20 expression as a statistically validated marker distinguishing these two histological grades.
MELK is a serine/threonine kinase involved in processes related to cell-cycle progression, apoptosis, DNA damage responses, and cancer-cell phenotypes (
15-
17). Increased MELK expression has been reported in aggressive breast cancer contexts. In the current study, MELK was selected as a shared hub gene and showed tumor-associated expression changes. However, as with CDC20, the Grade II-versus-Grade III comparison was not statistically significant (P = 0.820). Thus, the present findings support MELK as a candidate tumor-associated molecule but do not provide sufficient evidence to conclude that its expression differs according to histological grade in this cohort.
The miRNA analysis identified miR-192-5p and miR-17-5p as literature-supported candidates for CDC20 and MELK, respectively. Both miRNAs were selected from experimentally supported interactions shared by miRTarBase and TarBase and then prioritized based on the literature. In the clinical samples, miR-192 and miR-17 showed reduced tumor-associated expression relative to matched normal tissue. However, neither miRNA differed significantly between Grade II and Grade III tumors (P = 0.787 and P = 0.443, respectively). These results suggest that altered expression of these miRNAs may be associated with the tumor state but do not establish a grade-dependent expression pattern.
Table 4 shows that none of the tested correlations reached statistical significance. CDC20–MELK and miR-192–miR-17 showed positive but non-significant correlations, with P values of 0.081 and 0.052, respectively. The remaining correlations were weak or absent. Therefore, the current data should not be interpreted as evidence of coordinated expression or inverse gene–miRNA relationships. In particular, correlation analysis alone would not establish direct miRNA-mediated regulation even if a significant association had been observed.
The lack of statistically significant associations between biomarker expression and most clinicopathological variables further supports a cautious interpretation.
Table 5 shows no significant relationship with tumor type, pathological stage, DCIS, lymphovascular invasion, nipple involvement, family history, ER, PR, HER2, Ki-67, or age. Some variables contained highly imbalanced subgroups, such as ER and HER2 status, which limits the power of these comparisons. The non-significant findings should therefore be interpreted as an absence of evidence in this cohort rather than evidence that the biomarkers are unrelated to these characteristics.
The main strength of the study is its integration of public transcriptomic data, interaction-network analysis, external expression evaluation, miRNA-target databases, and experimental qRT-PCR validation. At the same time, several limitations should be considered. First, the validation cohort included only 30 tumors, limiting statistical power for grade comparisons and subgroup analyses. Second, the Grade I group in GSE10810 contained only two samples, making grade-specific conclusions involving Grade I exploratory. Third, the study did not include functional experiments such as miRNA mimic/inhibitor assays, luciferase reporter assays, gene knockdown, or rescue experiments. Consequently, the selected miRNAs cannot be considered experimentally confirmed regulators of CDC20 or MELK in the present system. Fourth, the absence of significant correlations in
Table 4 means that coordinated expression among the four markers cannot be inferred from the current data.
5.1. Conclusion
Integrated bioinformatics analysis identified CDC20 and MELK as shared hub genes in the Grade II and Grade III breast cancer networks, and external expression analysis supported increased expression of both genes in breast cancer. qRT-PCR evaluation of the clinical cohort further showed tumor-associated dysregulation of CDC20, MELK, miR-192, and miR-17 relative to matched adjacent normal tissue. However, the data presented in
Tables 4 and
5 do not support statistically significant correlations among the four biomarkers or statistically significant differences between Grade II and Grade III tumors. Accordingly, these molecules should be considered candidate tumor-associated biomarkers rather than validated grade-specific markers. Larger, independently characterized cohorts and functional studies are required to determine their prognostic and mechanistic significance.
5.2. Limitations
This study has several limitations. The bioinformatics analysis was based on a single public microarray dataset, and the Grade I subgroup included only two tumor samples; therefore, Grade I findings were considered exploratory. The hub-gene selection strategy relied on database-driven network analysis and may be influenced by the selected parameters and algorithms. In addition, the experimental validation cohort was relatively small, which may have limited the power to detect subtle differences and clinicopathological associations. Finally, although expression patterns were evaluated at the transcript level, no functional assays were performed to confirm the regulatory relationships between the selected miRNAs and their target genes. Further validation in larger, independent cohorts and mechanistic studies are needed.