Pathway analysis aids interpretation of large-scale gene expression data, but existing algorithms fall short of providing robust pathway identification. The method introduced here includes coexpression analysis and gene importance estimation to robustly identify relevant pathways and biomarkers for patient stratification.