Age-Specific Transcriptomic Signatures for Classification of Progression-Free Survival Outcomes in Luminal A Breast Cancer: An Integrative Machine Learning Approach.

Progression-free survival (PFS) is an important clinical endpoint in Luminal A breast cancer (BC), yet the molecular determinants of recurrence remain incompletely understood. This study aimed to identify recurrence-associated transcriptomic alterations across age-defined patient subgroups and evaluate their predictive utility using machine learning (ML). Gene expression profiles and clinical data from the METABRIC cohort were analyzed in premenopausal, postmenopausal non-geriatric, and geriatric patients with Luminal A BC. Differential expression analysis identified 32 significantly dysregulated genes, of which 15 genes with |log2FC| ≥ 2 were selected for detailed characterization. Among these, KMT2D, RFNG, IGF1, and CDKN2C exhibited consistent recurrence-associated expression patterns across patient subgroups, whereas ATM, RPTOR, and RICTOR displayed subgroup-dependent expression profiles. Feature selection analysis identified KMT2D as the most influential molecular predictor, while age, NPI, and tumor size were the most important clinical variables. The integrated clinicopathological-transcriptomic XGBoost model achieved the best predictive performance (AUC = 0.71; accuracy = 0.78) and outperformed baseline models based only on clinicopathological variables. Sensitivity analysis and independent external validation supported the robustness of the principal candidate biomarkers. These findings provide preliminary evidence supporting the investigation of transcriptomic biomarkers in combination with clinical variables for recurrence classification in Luminal A BC. However, further validation in independent cohorts is needed before their potential utility can be fully assessed.