Genomic analysis of radiosensitivity in breast cancer : Identifying pathological determinants and assessing genomic-adjusted radiation dose (GARD) for personalized dose escalation.

Introduction: Adjuvant radiotherapy improves recurrence-free survival in breast cancer, but intrinsic tumor radiosensitivity varies substantially, even within histologically similar subtypes. The radiosensitivity index (RSI), based on the expression of 10 genes, and the genomic-adjusted radiation dose (GARD) model enable personalized radiotherapy dosing. This study investigates the association between histological and molecular features and RSI, and quantifies the biological effect of radiation boost doses across conventional and hypofractionated regimens.

Materials And Methods: Transcriptomic RNA-seq data from 1284 breast cancer patients in The Cancer Genome Atlas Breast Invasive Carcinoma (TCGA-BRCA) cohort were analyzed. RSI was calculated using a rank-based model, and GARD was computed for multiple fractionation schemes, with or without integrated boosts. Univariate and multivariate linear models identified histological and molecular correlates of RSI. EPIC (estimating the proportions of immune and cancer cells) deconvolution was performed to estimate tumor purity and the immune/stromal cell composition. Analyses were restricted to samples with ≥ 50% tumor content. Independent validation was performed in the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) cohort (n = 1981), using microarray-based gene expression data.

Results: The median RSI in the TCGA cohort was 0.471 and was significantly lower in basal (p < 0.001) and luminal B (p < 0.001) subtypes, as well as in tumors with necrosis, inflammation, or high mitotic activity. These associations were replicated in the METABRIC validation cohort. Without a boost, 78.6% of the patients in the TCGA cohort would have achieved a GARD > 21 (associated with improved tumor control in retrospective series) with the 50 Gy/25 fractions regimen, compared to 64.8% for 40.05 Gy/15 fractions. The addition of an integrated boost significantly increased GARD values: 95.4% of patients receiving 64.4 Gy/28 fractions and 82.5% receiving 48 Gy/15 fractions achieved a GARD > 21. When stratified by molecular subtype, triple-negative breast cancer (TNBC) subtypes showed the greatest benefit from moderate dose escalation, with over 95% of these patients achieving GARD > 21 with a theoretical 53 Gy boost in 15 fractions. EPIC analysis revealed an inverse correlation between RSI and tumor cell content, and positive associations between RSI and specific immune or stromal components, highlighting the importance of tumor purity in interpreting RSI from bulk RNA data.

Conclusion: Our results support the biological relevance of RSI and GARD in breast cancer to personalize radiotherapy dose escalation in breast cancer patients and demonstrate their consistency across independent datasets and transcriptomic platforms. Tumor microenvironment composition significantly influences RSI estimation from bulk RNA-seq. Together, these findings support the implementation of personalized, biology-driven radiotherapy strategies, particularly for aggressive subtypes such as TNBC, and warrant prospective validation.