TITLE
Most reproduced "organoid-specific" cancer dependencies are consistent with culture-medium effects: a genome-wide medium-consistency audit of the 2026 organoid dependency catalogues

ABSTRACT (289 words)
Two 2026 Nature studies (Neiswender et al., Broad/NextGen; Herranz-Ors et al., Sanger) published genome-wide CRISPR dependency catalogues for patient-derived organoids, nominating organoid-specific dependencies as drug targets. Because organoid and conventional cultures differ in lipid supply, growth factors and other medium components, some dependency differences may reflect culture conditions. Neiswender et al. examined format effects genome-wide and selected medium effects using minipool and individual-gene assays. To our knowledge, catalogue-wide medium-consistency classifications and their corresponding fractions have not been reported. Here we apply prespecified medium- and format-consistency criteria to the organoid-vs-2D differential dependencies identified in our analysis of DepMap 26Q3 using organoid-medium models grown adherent (similar formulation family, no dome), dome-vs-adherent contrasts within organoid medium, and an unrelated serum-free lineage (CNS in Tumor Stem Media), with library (Avana vs Humagne-CD), lineage, and dynamic-range adjustment, on development patients with disjoint held-out-patient confirmation. Of 456 differential dependencies, 84–86% are classified medium-consistent under stated rules (75–82% under heteroscedasticity-robust SEs; bootstrap slope for adherent reproduction 0.90, 95% CI 0.82–0.99); one gene (LSM7) meets symmetric nominal criteria for format association and none do at 5% FDR. Among the Broad catalogue's published gains, 47% are medium-consistent overall and 87% of those that replicate under lineage adjustment; of Herranz-Ors et al.'s 18 headline genes, 8 of 16 present replicate — 5 medium-consistent (MVD, IDI1, MBTPS1 and MBTPS2 in the SREBP/sterol axis, plus PNKP) and 3 unresolved. Dependency losses behave similarly (84% medium-consistent); a relative ribosomal weakening survives global corrections in both assays — depletion kinetics under fewer doublings is a testable, unproven explanation. These are consistency classifications, not causal proof (adherent comparators: 6 of 9 lineages, mixed formulations). We release the per-gene table, with bootstrap flag frequencies, so any nominated organoid target can be checked before wet-lab investment.

SUBJECT AREA (suggested): Cancer Biology
LICENSE (suggested): CC-BY 4.0
ARTICLE TYPE: New Results / Technical Note resource
