Research use only. OSAConsensus is an integrated reanalysis resource for public obstructive sleep apnoea (OSA) and chronic intermittent hypoxia (CIH) transcriptomic data. Despite the name, it is not a clinical consensus statement or practice guideline and provides no diagnosis or treatment advice. Every program set carries its evidence-tier label; sets that failed cross-cohort or held-out validation are labelled as such next to the data.
Cross-species model fidelity
Pathway-layer fidelity (Step 6, main bridge layer): each animal (tissue × duration × background) combination's four-program ssGSEA effect profile (Hypoxia, Circadian, Inflammation, Oxidative stress) is correlated with the human chronic OSA blood reference profile. Gene-layer results are reported separately and labelled as the gene layer wherever the two are shown together.
The human reference profile is itself weak (no program at FDR < 0.05 in either backbone cohort), and with 4 programs the label-permutation space is only 24 (minimum achievable one-sided p ≈ 0.042): no combination is permutation-significant. Ranks are positions only — no model is “most human-like”.
Score-level statement: these are ssGSEA relative-enrichment effects. At the gene-mean logFC level the same sets are ≈ 0 (Step 7, item P5) — do not read as “hypoxia genes are on average down-regulated”.