Cross-Organ Transcriptomic Integration Reveals Cross-Disease Transcriptional Concordance and Shared Cell-State Patterns Between Metabolic Dysfunction-Associated Steatotic Liver Disease and Chronic Atrophic Gastritis

Background/Objectives: Integrating transcriptomic evidence across organs is limited by heterogeneity between cohorts, so reproducible cross-organ signals are usually reported as gene lists rather than as validated candidates. We therefore developed and applied a cross-organ integration framework in which discovery is separated from adjudication. Methods: Eight public bulk transcriptomic cohorts (six hepatic, two gastric) were combined by robust rank aggregation, and the shared candidates were carried into a pre-specified robustness audit combining leave-one-dataset-out recalculation, effect-size meta-analysis, and cell-composition correction. The candidate set was not re-selected after discovery, and the retained candidates were tested in two independent, non-discovery external cohorts. Results: Applied to metabolic dysfunction-associated steatotic liver disease and chronic atrophic gastritis, the framework returned four genes: CDHR2 and ANXA4 defined an epithelial pattern, IL32 an immune-associated pattern at the epithelial interface, and CADM2 a hepatic stellate-cell pattern; LGALS3 was retained separately as a composition-sensitive control. Every retained candidate kept its discovery-stage direction in both held-out cohorts. In the gastric replication cohort, where the test is run across ten pooled libraries, CDHR2 and CADM2 remained significant after Benjamini–Hochberg correction (both q = 0.020); for the other three candidates, the gastric contrast was directional only. In single-cell and single-nucleus data, the candidates resolved into three cell-state patterns that matched independent protein-level annotations, and exploratory Mendelian randomization returned no concordant, FDR-supported genetic evidence. Conclusions: Three of the five candidates (CDHR2, CADM2 and IL32) met all three prespecified criteria for the higher evidence tier and two (ANXA4 and LGALS3) the lower one. What the data establish is cross-disease transcriptional concordance along the gut–liver axis, interpreted through shared cell-state localization, rather than a shared etiology, a causal relationship, or a common biomarker. The audit framework is directly transferable to other cross-organ comparisons, and the five genes are reported as expression-supported candidates ready for experimental follow-up.