switchBox

"switchBox" is introduced as a response to the growing demand for a transparent, reproducible framework for developing genomic signatures amid concerns over the reliability of computational research in genomics. This need arises from the mixed success of genomic-based predictors in clinical settings. Despite a few success stories like MammaPrint, a predictive test for breast cancer, many genomic signatures fail to translate into clinically useful tools due to errors and a lack of statistical robustness.

"switchBox" addresses these issues by offering a fully reproducible analysis template for developing and evaluating genomic classifiers. It showcases this approach by re-analyzing the data used for MammaPrint, an FDA-cleared test based on a 70-gene expression signature for predicting breast cancer prognosis. "switchBox" not only reproduces the results of MammaPrint but also innovates by developing a simplified prognostic classifier that relies on just 16 genes from the original MammaPrint signature, achieving comparable accuracy in predicting 5-year disease-free survival.