Unraveling common molecular pathways in breast and ovarian cancer: a transcriptomic and bioinformatics approach
Article information
Abstract
Breast cancer is the most prevalent malignancy among women, while ovarian cancer, notably high-grade serous carcinoma, ranks as the most fatal gynecological cancer. Despite therapeutic advancements, both diseases face challenges related to metastasis and drug resistance. This study delved into the shared molecular mechanisms through a comparative bioinformatics approach. Transcriptomic datasets (GSE14407 for ovarian cancer and GSE33447 for breast cancer) were analyzed for differential expression, unveiling 105 commonly upregulated genes selected to aid downstream structure-based analyses. Protein-protein interaction networks were established, and hub genes were pinpointed via network topology analysis. Key regulators included ALDH1A3, ALDH1A1, AOX1, and ADH1B. Functional enrichment analysis highlighted involvement in metabolic adaptation, oxidative stress response, signaling, and tumor microenvironment remodeling, with pathways like retinol metabolism, glycolysis, cytochrome P450 drug metabolism, and tyrosine metabolism being enriched. Survival analysis further demonstrated that high expression of the ADH1B and ALDH1A3 hub genes consistently correlated with significantly enhanced overall survival, underscoring their prognostic relevance. These findings suggest that breast and ovarian cancers share crucial molecular pathways, pointing to common mechanisms of disease progression. Nonetheless, further experimental validation is essential to affirm these results and their therapeutic potential.
INTRODUCTION
Breast and ovarian cancers are the most common and deadliest malignancies impacting women globally. Breast cancer is the most frequently diagnosed cancer and a primary contributor to female cancer-related deaths. Ovarian cancer, while less prevalent, stands out as the most lethal gynecologic malignancy due to its aggressive nature and late detection [1,2].
Both cancers share several genetic and hormonal risk factors, most notably BRCA1 and BRCA2 mutations, which significantly increase the lifetime susceptibility to both diseases [3]. Epidemiological and molecular evidence suggest overlapping etiological pathways, such as estrogen exposure, reproductive factors, and defects in DNA repair mechanisms, contribute to their pathogenesis [4,5]. Early menarche, late menopause, nulliparity, and obesity increase risk, while parity and breastfeeding have protective effects [6].
At the molecular level, breast cancer is classified into main subtypes: luminal A/B, HER2-enriched, and triple-negative, each characterized by specific genomic alterations (e.g., PIK3CA, ERBB2, and TP53) [7]. Ovarian cancer, mainly a high-grade serous carcinoma, often shows TP53 mutations and BRCA-related homologous recombination deficiencies [8]. Although originating from different tissues, these cancers share common pathways related to DNA repair, metabolism, and regulation of oxidative stress.
Recent studies have indicated that breast and ovarian cancers have common molecular biomarkers and regulatory networks, which supports the need for an integrated analysis to reveal shared oncogenic mechanisms. Discovering these genes and pathways could enhance our comprehension of cross-cancer biology, identify potential therapeutic targets, and guide strategies for early cancer detection and drug repurposing [9].
The present study aimed to identify common genetic biomarkers between breast and ovarian cancers through microarray-based differential expression analysis, protein-rotein interaction networks, and functional enrichment analyses. By comparing the transcriptomic profiles of both cancer types, we aimed to elucidate common molecular mechanisms driving their co-occurrence and guide the development of targeted therapeutic strategies. Prioritizing upregulated genes from a therapeutic standpoint also enhances subsequent structure-based analyses, such as molecular docking or drug repurposing studies, as overexpressed proteins present more promising pharmacological targets for inhibition.
MATERIALS AND METHODS
1. Research methodology
The schematic workflow summarizing the bioinformatics pipeline is illustrated in Figure 1.
Graphical flowchart of the research methodology. BP: biological process, CC: cellular component, DEGs: differentially expressed genes, FC: foldchange, GO: gene ontology, KEGG: Kyoto Encyclopedia of Genes and Genomes, MCC: maximal clique centrality, MF: molecular function, MNC: maximum neighborhood component, PPI: protein-protein interaction.
2. Data retrieval
Microarray datasets for ovarian cancer (GSE14407) and breast cancer (GSE33447) were downloaded from the NCBI Gene Expression Omnibus (GEO) [10], utilizing the GEOquery package (v2.68.0) in R Studio (v4.3.1) (https://posit.co/download/rstudio-desktop/). Both datasets were obtained from the Affymetrix Human Genome U133 Plus 2.0 Array (GPL570) platform, comprising 24 ovarian samples (12 normal ovarian surface epithelium and 12 epithelial ovarian carcinoma) and 16 breast samples (eight normal and eight tumor tissues), without additional human sample collection. Since this study exclusively relied on publicly available datasets, Institutional Review Board (IRB) approval was not required. This study did not involve human participants or patient data. Only publicly available datasets obtained from the GEO was analyzed. Therefore, IRB approval and informed consent were not required.
3. Data processing and differential expression analysis
The raw intensity data were imported into R and processed following the limma microarray workflow [11]. Signal distributions were examined, and log2 transformation was applied as needed. Background correction and quantile normalization were carried out using the normalizeBetweenArrays() function to eliminate technical variations across the arrays.
Feature metadata from the latest GPL570 annotation file were utilized to map the probe identifiers to official gene symbols. Probes without valid gene annotations were excluded. In cases where genes were targeted by multiple probes, the expression values were averaged with the dplyr package in R to derive a singular representative measurement per gene. Quality control involved the utilization of boxplots and principal component analysis to evaluate data consistency and identify potential outliers, of which none were detected. Given that each GEO dataset stemmed from a distinct experimental batch, batch-effect correction was unnecessary. The preprocessing and normalization steps were executed using the limma microarray analysis framework [12].
Differential expression analysis was performed between the tumor and normal groups utilizing the limma linear model framework (lmFit, contrasts, fit, eBayes). Significance was determined as an adjusted P-value<0.05 (Benjamini-Hochberg false discovery rate [FDR]) and |log2 fold-change|≥ 1. To balance sensitivity and control for false discoveries due to the limited sample sizes (12 vs. 12 and 8 vs. 8), these thresholds were chosen. The resulting normalized gene-level matrices were exported as tab-delimited files (expression_genes.tsv) to ensure reproducibility and facilitate subsequent integrative analyses.
Overlapping common upregulated genes between breast and ovarian cancers were identified using the Venny online Venn diagram tool (https://bioinfogp.cnb.csic.es/tools/venny) [13].
4. Protein-protein interaction network construction, analysis, and hub gene identification
The identified shared genes, determined through Venny, were subsequently mapped to the STRING database (https://string-db.org/) to extract protein-protein interactions [14]. The resulting protein-protein interaction networks were then visualized and analyzed utilizing Cytoscape (https://cytoscape.org/). Topological properties such as node degree, betweenness centrality, stress centrality, and closeness centrality were evaluated employing the CytoHubba plugin [15]. Hub genes were ranked based on their degree values, leading to the identification of the top 10 genes as key candidates. Module analysis was conducted using MCODE (v2.0.0) to further confirm the significant clusters within the network [16].
5. Gene enrichment analysis
Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were conducted using DAVID v6 (https://www.genome.jp/kegg/pathway.html). Gene ontology (GO) categories comprised biological processes (BPs), cellular components (CCs), and molecular functions (MFs) [17].
6. Survival analysis
To determine the prognostic significance, survival analysis of the selected hub genes was conducted using the GEPIA2 web server, which is freely accessible online (http://gepia2.cancer-pku.cn/#survival). This tool generates overall survival plots from The Cancer Genomic Atlas (TCGA) datasets. Each hub gene was assessed individually to establish its correlation with patient survival outcomes. The corresponding Kaplan-Meier curves and hazard ratios were analyzed to evaluate their potential clinical relevance [18].
RESULTS
1. Identification of differentially expressed genes
The ovarian cancer (OC) dataset GSE14407 comprised 24 samples (12 normal and 12 tumor), while the breast cancer (BC) dataset GSE33447 comprised 16 samples (eight normal and eight tumor). Differential expression analysis was conducted on both datasets, and the outcomes were depicted through volcano plots in Figure 2A and 2B. A Venn diagram of the upregulated genes identified 105 differentially expressed genes that were commonly expressed during both OC and BC progression, as shown in Figure 2C.
Volcano plots of datasets (A) GSE14407 (breast cancer) and (B) GSE33447 (ovarian cancer) showing distribution of DEGs. Red, blue, and grey colors indicate up-regulated genes, down-regulated genes, and non-significant genes. (C) Common upregulated genes between ovarian cancer and breast cancer using a Venn diagram.
2. Protein-protein interaction construction, analysis, and identification of hub genes
A protein-protein interaction network of the 105 commonly upregulated genes in BC and OC was generated using the STRING database, with a high confidence threshold of 0.7, indicating biologically relevant or experimentally validated interactions. The resulting network comprised 101 nodes (genes) and 29 edges (interactions) (Fig. 3A). To identify key regulatory genes in the protein-protein interaction network, we utilized the CytoHubba plugin in Cytoscape, employing two centrality-based algorithms: maximal clique centrality (MCC) (Fig. 3B) and maximum neighborhood component (MNC) (Fig. 3C).
(A) PPI networks generated using STRING with 101 nodes and 29 edges. Top 10 ranked hub genes network obtained using the Cytoscape plugin CytoHubba (B) MCC and (C) MNC. MCC: maximal clique centrality, MNC: maximum neighborhood component, PPI: protein-protein interaction.
The MCC algorithm ranked AOX1, ADH1B, ADH1C, ALDH1A1, and ALDH1A3 as the top five hub genes, indicating their presence in tightly interconnected cliques within the network. The MNC algorithm also highlighted AOX1, ADH1B, ADH1C, ALDH1A1, and ALDH1A3, along with MAOB, SNCA, CAV1, CAVIN2, and CTNNAL1, based on neighborhood connectivity. Notably, AOX1, ADH1B, ADH1C, ALDH1A1, and ALDH1A3 were consistently detected by both algorithms, suggesting their central roles in maintaining network structure and function. In contrast, monoamine oxidase B (MAOB) and synucleic alpha (SNCA) were identified exclusively by the MNC, reflecting broader connectivity but weaker clique participation than the core hub set. Overall, AOX1, ADH1B, ADH1C, ALDH1A1, ALDH1A3, and MAOB emerged as robust hub genes (Table 1), were consistently prioritized across multiple centrality measures, and potentially served as key drivers of the biological processes represented by the network.
Ranking of hub genes based on MCC and MNC scores using CytoHubba
3. Gene enrichment analysis
Functional annotation of the identified hub genes was conducted using DAVID to investigate their biological significance. The analysis demonstrated that the genes were notably enriched in metabolic pathways and retinol metabolism, as revealed by KEGG pathway enrichment (P=2.1×10-4, FDR=1.3×10-3). Furthermore, the hub genes were significantly linked to the oxidoreductase activity category (P=3.7×10-7), emphasizing their crucial involvement in redox homeostasis (Table 2).
DAVID enrichment analysis of hub genes showing significant KEGG pathways and GO terms
GO classification further supported these findings, indicating enrichment in oxidoreductase activity, nicotinamide adenine dinucleotide (NAD) binding, and zinc ion binding within molecular function terms. The hub genes were found to be involved in retinoid and alcohol metabolic processes in the biological process category, with their localization primarily in the cytoplasm and plasma membrane according to cellular component analysis (Table 2).
4. Survival analysis of hub genes
The TCGA-based validation using survival analysis revealed a significant protective association with overall survival in breast and ovarian cancers for all six hub genes. Each GEPIA2 Kaplan-Meier plot (Fig. 4) demonstrated that high expression of ALDH1A3, ALDH1A1, ADH1B, ADH1C, AOX1, and MAOB was significantly associated with enhanced overall survival in OC and BC. ALDH1A3 exhibited one of the strongest effects, with a hazard ratio (HR) of 0.53 (P=2.5×10-10) (Fig. 4D), while ALDH1A1 reduced mortality risk by approximately 30% (HR, 0.70; P=0.00044) (Fig. 4C). The ADH1 family members also conferred a survival benefit, with ADH1C (HR, 0.63; P=4×10-6) (Fig. 4B) and ADH1B (HR, 0.57; P=2.6×10-8) (Fig. 4A). AOX1 expression corresponded to a 31% risk reduction (HR, 0.69; P=0.00019) (Fig. 4E), while MAOB exhibited the most protective effect (HR, 0.45; P=1×10-14) (Fig. 4F). These results were consistent with the enrichment findings, supporting the role of these metabolic genes in redox regulation and tumor suppression. In each plot, the high-expression group demonstrated significantly better survival compared to the low-expression group, indicating a protective prognostic role.
(A-F) Overall survival analysis of the six common hub genes in breast cancer and ovarian cancer. HR: hazard ratio.
DISCUSSION
In this study, we identified shared molecular signatures between BC and OC using a comparative bioinformatics approach. Differential expression analysis revealed that 105 genes were upregulated. Upregulated genes are more likely to produce overexpressed proteins suitable for pharmacological inhibition. These results lay the groundwork for subsequent structure-based and drug-repurposing studies, as well as suggest potential cross-cancer mechanisms.
CytoHubba consistently ranked AOX1, ADH1B, ADH1C, ALDH1A1, and ALDH1A3 (Table 1) as central nodes using both MCC and MNC algorithms. Additionally, MNC highlighted MAOB and SNCA. These genes primarily function in metabolic and redox-related functions, indicating a shared mechanism between BC and OC involving disrupted cellular metabolism. For example, aldehyde oxidase 1 (AOX1) plays a role in xenobiotic metabolism and the generation of reactive oxygen species (ROS). Abnormal AOX1 activity has been associated with oxidative stress, a key feature of tumor progression [17].
Elevated ALDH1A1/ALDH1A3 activity is linked to cancer stemness and therapy resistance. Dysregulated alcohol dehydrogenase (ADH) and aldehyde dehydrogenase (ALDH) expression is commonly observed in cancer, serving as metabolic regulators and stemness markers, particularly ALDH1A1 in breast cancer [19,20]. MAOB, exclusively identified in MNC, controls neurotransmitter metabolism and aids in maintaining ROS balance, thereby connecting mitochondrial dysfunction to cancer pathogenesis [21]. SNCA, traditionally associated with neurodegenerative disorders, has been implicated in tumorigenesis by influencing vesicular trafficking and resistance to apoptosis [22].
KEGG analysis indicated significant enrichment in retinol metabolism, general metabolic pathways, tyrosine metabolism, cytochrome P450 (CYP)-mediated drug metabolism, glycolysis/gluconeogenesis, and fatty acid degradation. Notably, retinol metabolism emerged as the most significantly enriched pathway (P=1.8×10-8), aligning with the functions of ALDH1A1 and ALDH1A3 in retinoic acid biosynthesis. Retinoic acid signaling plays a crucial role in regulating cell differentiation, proliferation, and apoptosis, with its dysregulation being a key feature of tumorigenesis [23]. In breast cancer, ALDH-positive cells exhibit cancer stem cell-like properties, while in ovarian cancer, retinoid signaling is linked to chemoresistance.
General metabolic pathways (P=2.1×10-4) exemplify metabolic reprogramming, characterized by shifts in glucose, fatty acid, and amino acid metabolism that sustain tumor growth and survival [24]. Tyrosine metabolism (P=6.9×10-7) and tryptophan metabolism (P=2.4×10-2) further signify disruptions in amino acid utilization. These pathways play crucial roles in energy production, redox balance, and immune evasion. Specifically, altered tryptophan metabolism via the kynurenine pathway has been associated with immune suppression in the tumor microenvironment [25]. Drug metabolism involving CYP (P=5.9×10-6) highlights potential interactions between metabolic enzymes and therapeutic agents. Dysregulation of CYP-mediated detoxification may contribute to chemoresistance [26], a significant challenge in both OC and BC. Glycolysis/gluconeogenesis and fatty acid degradation demonstrate the well-documented metabolic rewiring of cancer cells, often referred to as the Warburg effect, where enhanced glycolysis supports rapid cell proliferation [27].
GO analysis further clarified the molecular roles of hub genes. Within the MF category, there was significant enrichment in oxidoreductase activity, specifically in enzymes that act on alcohol, aldehyde, or NAD(P)+ donors. These results align with the known functions of the ADH and ALDH gene families in redox regulation and detoxification [28].
Additionally, enhancement of retinol dehydrogenase activity and NAD binding reflects their significance in retinoid signaling and cofactor metabolism [23]. In terms of BP, hub genes exhibited enrichment in retinol and retinoic acid metabolic processes, supporting the KEGG findings and emphasizing the pivotal role of retinoid pathways in cancer development by influencing cell fate decisions and stemness [23]. Regarding CC, the genes were predominantly localized in the cytosol and plasma membrane, aligning with their enzymatic and metabolic functions. Notably, enrichment in ciliary basal body and cilia-related terms suggests potential links between ciliary dysfunction and tumorigenesis, as cilia regulate key signaling pathways such as Hedgehog and Wnt, which are frequently dysregulated in cancer.
1. Limitations and future directions
Based on the shared transcriptomic signatures identified in this study, we suggest that breast and ovarian cancers share common metabolic and oxidative stress pathways regulated by ALDH1A1, ADH1B, and AOX1. These genes may serve as central regulators of retinoid metabolism and redox balance, thereby enhancing cell survival and therapy resistance in both cancer types. This hypothesis can be experimentally validated through functional assays, such as gene knockdown or overexpression in breast and ovarian cancer cell lines, to evaluate their impact on ROS levels, proliferation, and drug sensitivity. Additionally, survival and expression correlation analyses using larger patient cohorts (e.g., The Cancer Genome Atlas Breast Invasive Cancer [TCGA-BRCA] and The Cancer Genome Atlas-Ovarian Cancer [TCGA-OV]) could assess the prognostic significance of these hub genes. Confirming these associations will bolster the mechanistic connection between shared metabolic reprogramming and clinical outcomes.
CONCLUSIONS
This study identified 105 commonly upregulated genes shared between breast and ovarian cancers, revealing convergent molecular mechanisms. Hub genes such as ALDH1A1, ALDH1A3, ADH1B, AOX1, ADH1C, and MAOB were enriched in retinol metabolism and oxidative stress regulation, pathways known to influence tumor survival and chemoresistance. Increased ALDH1A1/ALDH1A3 activity is linked to cancer stemness and therapy resistance, while altered retinoid metabolism can maintain undifferentiated drug-tolerant cell populations. These results imply that metabolic and redox reprogramming present shared vulnerabilities in both cancer types. These genes are promising targets for structure-based modeling and molecular docking studies to investigate their therapeutic potential and drugability.
Notes
ACKNOWLEDGEMENTS
All authors contributed. This study was conducted independently.
CONFLICT OF INTEREST
The authors report no conflict of interest.
FUNDING
This research was conducted independently and did not receive any specific funding from public, commercial or not-for-profit organizations.