CROSS-PLATFORM CONCORDANCE ANALYSIS REVEALS ROBUST AND CLINICALLY RELEVANT GENE SIGNATURES IN BREAST CANCER

Abstract

Abstract Background: Breast cancer is a molecularly heterogeneous disease characterized by diverse transcriptional programs across tumor subtypes. RNA sequencing (RNA-seq) has become a central tool for profiling these alterations; however, variability in data processing pipelines can introduce inconsistencies in downstream analyses, limiting the identification of robust and reproducible gene signatures. Methods: We implemented a three-phase integrative framework to identify high-confidence gene signatures in breast cancer. The framework comprised: (i) a discovery phase involving differential expression analysis across two independent RNA-seq workflows; a STAR-counts pipeline applied to a balanced TCGA-BRCA cohort (30 tumor vs. 30 normal samples) processed via TCGAbiolinks, and a Salmon-based pseudo-alignment pipeline applied to an independent SRA-derived dataset (3 tumor vs. 3 normal samples) processed from raw sequencing reads; (ii) a cross-dataset concordance phase identifying consensus upregulated genes through gene symbol intersection analysis; and (iii) a validation and clinical characterisation phase comprising large-scale validation in the full TCGA-BRCA cohort (n = 1,093 tumors, 112 normal samples), functional enrichment, PAM50 subtype stratification, survival analysis, and cancer cell line expression profiling. A multi-criteria scoring framework was applied to prioritize potential drug targets. Results: A total of 666 genes were consistently upregulated across both discovery datasets, indicating strong cross-platform concordance. Of these, 502 genes (93%) were validated as upregulated in the full TCGA cohort, demonstrating high reproducibility. Functional enrichment analyses revealed dominant involvement in cell cycle regulation, chromosomal segregation, and DNA repair pathways. Subtype analysis showed a marked enrichment of these genes in Basal-like and HER2-enriched tumors, highlighting their association with more aggressive disease phenotypes. Survival analysis identified a subset of genes significantly associated with poor overall survival, while cell line validation confirmed expression in breast cancer models. Integration of these features enabled the prioritization of candidate genes with potential therapeutic relevance. Conclusion: This study demonstrates that integrating two independent RNA-seq analytical workflows (a STAR-counts pipeline applied to a balanced TCGA-BRCA discovery cohort and a Salmon-based pipeline applied to an independent SRA-derived dataset) enables the identification of robust and clinically meaningful gene signatures in breast cancer. The high concordance observed across these independent datasets underscores the reproducibility of the identified signals, while downstream analyses highlight their biological and translational relevance. This framework provides a systematic approach for prioritizing candidate biomarkers and therapeutic targets in cancer transcriptomics.