The NCBI-Gene Expression Omnibus (GEO) [8] is a publicly accessible database that houses microarray data. It is extensively utilized for gene expression datasets and platform records. For this BC, we obtained gene expression datasets from NCBI-GEO and analyzed them using GEO2R online tool. Differentially expressed genes (DEGs) were identified using Benjamini & Hochberg’s false discovery rate (FDR), considering genes with p < 0.05 and logFC > 2.0 as significantly upregulated, data available in the Supplementary material.

We constructed protein-protein interaction (PPI) networks using STRING [9] and analyzed them in Cytoscape (v3.10.2) using Molecular Complex Detection (MCODE) and CytoHUBba plugins. We selected genes with at least three overlapping algorithms to identify highly upregulated biomarker (HUB) genes.

In total, 456 upregulated DEGs were extracted from the four datasets, and were utilized to construct the PPI network for the Asian population, and 2,950 upregulated DEGs were extracted from the three datasets, and were utilized to construct the PPI network for the European population.

Further analysis and visualization of the network were conducted using Cytoscape software (v3.10.2) [10]. Respective STRING networks were analyzed on Cytoscape using MCODE and CytoHUBba plugins. The networks were analyzed by MCODE and densely connected regions (clusters) within large PPI networks were identified. CytoHUBba was used to identify important nodes and subnetworks [11]. All 12 algorithms in cytoHUBba were used for analysis, namely: Degree Centrality, Betweenness Centrality, Closeness Centrality, Stress Centrality, Eccentricity, Radiality, BottleNeck, Edge Percolated Component (EPC), Maximum Neighborhood Component (MNC), Density of MNC (DMNC), Clustering Coefficient, and Maximal Clique Centrality (MCC).

Gene set enrichment was performed using Enrichr-KG and GeneMANIA online bioinformatic tools to explore associations with known metabolic pathways, particularly cholesterol metabolism, adipocytokine signaling, and AMP-activated protein kinase (AMPK) regulation. Pathways with an adjusted p-value of less than 0.005 were considered [12, 13].

For BC, four GEO datasets with accession numbers GSE29044, GSE15852, GSE89116, and GSE61304 for the Asian population and three GEO datasets with accession numbers GSE42568, GSE21422, and GSE29431 for the European population were retrieved from the freely accessible NCBI-GEO database. GSE15852 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE15852 (Malaysia), GSE29044 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE29044 (Riyadh, Saudi Arabia), GSE89116 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE89116 (New Delhi, India), and GSE61304 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE61304 (Singapore) datasets were chosen for BC analysis of Asian population and GSE29431 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE29431 (Barcelona, Spain), GSE21422 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE21422 (Berlin, Germany), and GSE42568 https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE42568 (Dublin, Ireland) were chosen for the European population. Samples were taken from patients of different age groups and at different stages of cancer progression, DGE analysis results of these datasets provide HUB genes whose expression in BC could be dependent on environmental factors, diet, genetic history, and lifestyle disorders. Set of images in Figure 2a graphical representation of DEGs in Asian datasets, and set of images in Figure 2b graphical representation of DEGs in European datasets.

In total, 2,950 upregulated DEGs were extracted from the three datasets, and were utilized to construct the PPI network for the European population and 456 upregulated DEGs were extracted from the four datasets, and were utilized to construct the PPI network for the Asian population. The PPI networks obtained from STRING [9] are shown in supplementary figures: Figures S1–7.

The common genes were shortlisted using Bioinformatics & Evolutionary Genomics Venn diagram generator tool [14]. In selected BC datasets, we selected the genes which had minimum of 3 common algorithms in that particular dataset, for each (intra-dataset comparison), files are attached in the Supplementary material under CytoHUBba algorithms and CytoHUBba analysis. Followed by, every dataset’s selected gene list to check for common genes within them (inter-dataset comparison) to get HUB genes. The lists are mentioned in Table 1.

Cluster of differentiation 36 (CD36), leptin (LEP) were found to be common genes among GSE29044, GSE89116, and GSE15852 datasets and Singapore consisted of cyclin-dependent kinase 1 (CDK1), aurora kinase A (AURKA), DNA topoisomerase II alpha (TOP2A), and baculoviral IAP repeat containing 5 (BIRC5) (Asian). Caveolin-1 (CAV1), insulin-like growth factor 1 (IGF1), CD36, apolipoprotein B (APOB), peroxisome proliferator-activated receptor gamma (PPARG), LEP were found to be common genes among GSE42568, GSE21422 and GSE29431 datasets (European).

Venn diagram of GSE29044, GSE89116, GSE15852, and GSE61304 datasets (Asian population) and Venn diagram of GSE42568, GSE21422, and GSE29431 datasets (European population) is shown in Figure 3.

Table 2 contains tabulated data of Venn diagram of Asian and European Population. From Venn diagram Asian results we observe that there are no genes common between all four Asian datasets but CD36 and LEP were found as HUB genes between GSE29044, GSE89116, and GSE15852 datasets, and CAV1 and perilipin 1 (PLIN1) were found as secondary HUB genes common to GSE89116 and GSE15852. The Singapore data set consisting specifically of breast adenocarcinoma did not have any common HUB gene, with the other selected datasets as the other datasets were not inclusive only of breast adenocarcinoma. From Venn diagram European results we observe that CAV1, IGF1, CD36, APOB, PPARG, and LEP are common HUB genes found in European population.

CD36 and LEP were found to be common HUB genes in both populations, while CAV1 and PLIN1 were prevalent in Asian datasets, and CAV1, IGF1, APOB, and PPARG were identified in European datasets (Table 1).

A collective HUB gene analysis across all datasets reinforced the significance of CD36, LEP, and PPARG as key molecular regulators in BC progression.

This study highlights CD36 and LEP as promising therapeutic targets, warranting further wet lab validation and clinical trials. Additionally, integrating multi-omics approaches and AI-driven predictive models could refine precision oncology frameworks.

BC exhibits significant molecular heterogeneity influenced by genetic, dietary, environmental, and epigenetic factors. This study identified CD36 and LEP as universal HUB genes consistently upregulated across both Asian and European BC datasets, indicating their potential as global diagnostic and prognostic biomarkers. Additionally, CAV1, IGF1, APOB, and PPARG were identified as key regulatory genes in European populations, while PLIN1 and CAV1 were prominent in Asian populations, suggesting region-specific molecular variations that may influence tumor progression and therapeutic response.

Functional enrichment analysis highlighted the involvement of these genes in critical pathways, including cholesterol metabolism, adipocytokine signaling, AMPK regulation, and lipid metabolism, emphasizing their role in BC pathophysiology. The observed differences in gene expression patterns underscore the necessity for personalized medicine approaches, incorporating molecular profiling to optimize early detection, risk stratification, and targeted treatment strategies.

Clinically, CD36 and LEP may serve as potential biomarkers for non-invasive liquid biopsy screening, enabling early detection and tumor stratification. Their role in metabolic regulation also suggests potential therapeutic applications, with CD36 inhibitors and LEP antagonists warranting further exploration in obesity-associated BC. Furthermore, PPARG agonists and lipid metabolism modulators may serve as promising candidates for targeted interventions, particularly in hormone receptor-positive BC subtypes. Given the ethnic and regional variations in HUB gene expression, incorporating population-specific molecular profiling into treatment regimens could enhance therapeutic efficacy and minimize adverse effects.

Future research should focus on large-scale cohort validation and experimental studies to confirm the clinical utility of these biomarkers and evaluate their therapeutic potential. Additionally, the integration of multi-omics approaches, AI-driven predictive models, and metabolic interventions could refine precision oncology strategies, ensuring more effective, patient-centered treatment paradigms. CD36 and LEP emerge as promising targets for advancing BC diagnostics, prognostics, and therapeutics, paving the way for more precise and individualized management strategies in breast oncology.