A Versatile and Upgraded Version of the LundTax Classification Algorithm Applied to Independent Cohorts

Abstract

Stratification of cancer into biologically and molecularly similar subgroups is a cornerstone of precision medicine. The Lund Taxonomy classification system for urothelial carcinoma aims to be applicable across the whole disease spectrum including both non-muscle-invasive and invasive bladder cancer. A successful classification system is one that can be robustly and reproducibly applied to new samples. However, transcriptomic methods used for subtype classification are affected by analytic platform, data preprocessing, cohort composition, and tumor purity. Furthermore, only limited data have been published evaluating the transferability of existing classification algorithms to external data sets. In this study, a single sample classifier was developed based on in-house microarray and RNA-sequencing data, intended to be broadly applicable across studies and platforms. The new classification algorithm was applied to 10 published external bladder cancer cohorts (n = 2560 cases) to evaluate its ability to capture characteristic subtype-associated gene expression signatures and complementary data such as mutations, clinical outcomes, treatment response, or histologic subtypes. The effect of sample purity on the classification results was evaluated by generating low-purity versions of samples in silico. The classifier was robustly applicable across different gene expression profiling platforms and preprocessing methods and was less sensitive to variations in sample purity.