PathwayBench v2: Rank-based pathway scoring methods can invert biological signal under gene competition

Companion data and code for Alakwaa et al. (submitted to Nature Methods, April 2026). Benchmark of five pathway activity scoring methods (ssGSEA, GSVA, z-score, AUCell, UCell) across seven disease datasets totaling 4,005,940 cells, 618 donors, and four tissues. The deposit contains the manuscript and supplementary information, 14 figures, 7 pseudobulk count matrices with metadata (sum and mean aggregation), curated ground-truth pathway annotations, complete evaluation results, and a Snakemake pipeline for full reproducibility. Central finding: rank-based methods (AUCell, UCell) can produce sign-inverted pathway scores when non-pathway genes compete for top-rank slots, demonstrated by simulation and confirmed in CKD kidney ECM remodeling real data.