TumorArchetypeR

This package provides a complete workflow to discover robust tumor archetypes. It automates the following key steps: Scoring: Computes enrichment scores for gene signatures using ssGSEA or loads pre-calculated scores. Normalization: Optionally normalizes scores by tumor purity and scales signatures to ensure equal contribution. Dimensionality Reduction: Performs PCA on the enrichment scores. Clustering: Applies Louvain clustering across a user-defined grid of parameters (e.g., different variance cutoffs and k-neighbors). Characterization: Evaluates each clustering result using internal metrics, cluster stability, and association with clinical variables. Finally, the package helps you select an optimal clustering result and train a robust single-sample classifier that can be used to apply the derived subtypes to new datasets.