Abstract
Intrinsic molecular subtyping (IS) of breast cancer is fundamental for understanding disease biology and guiding personalized treatment. While IS methods are standardized in clinical practice, their research implementations vary, leading to inconsistencies and reduced reproducibility. We introduce BreastSubtypeR, an R/Bioconductor package that unifies multiple published IS classifiers into a single, reproducible framework, enabling systematic cross-method comparison and robust cohort-aware selection. Its core feature, AUTO mode, quantitatively evaluates cohort composition and selectively activates classifiers whose assumptions are met, reducing user bias and improving robustness across skewed or subtype-specific cohorts. The package also provides a unified, method-specific normalization pipeline, optimized probe-to-gene mapping, and an intuitive local Shiny app (iBreastSubtypeR) for non-programmers. By standardizing method selection, preprocessing, and mapping within a Bioconductor workflow, BreastSubtypeR improves reproducibility and accessibility, and addresses an urgent need for assumption-aware IS in breast cancer research. Availability and implementation: The BreastSubtypeR package and its companion R Shiny application, iBreastSubtypeR, are freely available through Bioconductor (https://doi.org/10.18129/B9.bioc.BreastSubtypeR). The complete source code is also hosted on GitHub (https://github.com/JohanHartmanGroupBioteam/BreastSubtypeR). A citable archived snapshot of the version used (v1.1.3) is available on Zenodo (https://doi.org/10.5281/zenodo.17085316). The software is released under the GPL-3 license. Comprehensive documentation and an example dataset are provided to facilitate user adoption.