Cell-death regulation in patient tumors, 3D tumoroids and 2D cell lines: source data for Figure 6

This dataset contains derived results from a reanalysis of previously published transcriptomic datasets and functional genetic screens. It supports Figure 6 of “Cell-death programs and responses to immune pressure across epithelial tumor models”. The original samples, sequencing data and experimental screens were generated by the research groups cited below. This deposition provides the comparative analysis, derived tables and figure exports. The expression comparison comprises 804 profiles across six anatomical sites: 20 healthy epithelial profiles, 231 malignant epithelial profiles, 152 3D dome models and 401 2D adherent models. These are separate cohorts matched by anatomical site, not paired patient tumors and derived cultures. The analysis uses a common universe of 18,612 genes and 12 selected mechanistic groups containing 67 genes: 41 from a broader 108-gene candidate panel and 26 additional genes involved in antigen presentation and immune interactions. Healthy epithelial data derive from Tabula Sapiens 2.0. Malignant epithelial data derive from the atlas of Kang et al. (2024), using the processed matrices in Zenodo record 10651059, version v5. Traditional 2D model data derive from DepMap/CCLE; 3D model expression and annotations derive from the DepMap file set “NextGen Model Manuscript 2026” associated with Neiswender et al. The functional analysis uses six CRISPR selection arms from Watterson et al. (Supplemental Table S7), one pooled HeLa arm from Zhou et al. (supplementary materials 4 and 5), the HT-29 siRNA candidate list of Woznicki et al. (Supplementary Table S1), and the TCR-selection candidate list of Patel et al. (Supplementary Information Table 1). SOURCE_DATASETS.tsv identifies the source components, publication DOIs and data locations. References are provided below and in REFERENCES.md. The reanalysis compares expression ranks relative to matched control genes, aggregates expression scores across anatomical sites, estimates conditional cluster-bootstrap intervals and descriptive random-gene reference bands, and compares functional-screen summaries and published candidate lists. Expression scores describe relative transcript representation; they do not measure pathway activity or clinical resistance. The selected gene groups are mechanistic collections rather than a validated resistance signature. The archive includes profile scores, expression summaries, functional-screen summaries, gene-set definitions, provenance tables, data dictionaries, file checksums, the composite Figure 6 in PDF and PNG, three standalone panels, the figure legend and methods. The MLKL/PGAM5 sensitivity analysis reports both Monte Carlo and exact finite-pool reference results. These descriptive analyses do not establish a cell-death modality. The background-control correction introduced in version 1.1.0 excludes every scored gene from its own 400-gene control pool. All 48 expression reference bands were recomputed; five of 24 normalized contrasts changed by at most 0.04 reference-band units. The 72 analyzed estimates, their conditional bootstrap intervals and all Figure 6 reference-band classifications were unchanged. The correction report, control assignments and 127-file primary-input inventory are included. Version 1.1.1 updates documentation, source attribution and citation metadata. All 51 numerical and provenance tables, the composite figure and the three standalone panels are byte-identical to version 1.1.0. The companion software reproduces the figure from deposited profile-level scores and derived screen summaries. Primary-data processing uses separately obtained source files identified in SOURCE_DATASETS.tsv and docs/PRIMARY_INPUTS.md. Primary expression matrices and raw sequencing reads remain with their original repositories. The exact quarterly identifier of the traditional DepMap input was not recorded; the input inventory retains filenames, sizes and reported checksums. Original resources retain their own provenance and reuse conditions. Analysis and plotting software: https://github.com/alexeysakhalin/tumor-model-comparison-patient-3d-2d/releases/tag/v1.1.1