Continuous multi-omics pathway enrichment analysis resolves hidden functional heterogeneity.

Pathway enrichment analysis is essential for extracting biological insights from complex omics datasets, yet existing methods suffer from critical limitations: excessive false discoveries, arbitrary significance thresholds, poor handling of multi-omics data, and inability to model gene dependencies. We present JOANA (Joint continuous multi-Omics enrichment ANAlysis), a novel Bayesian framework for pathway analysis with three key contributions. First, JOANA enables high specificity through continuous probabilistic modeling of significance scores using Beta mixture distributions, eliminating arbitrary thresholds while maintaining sensitivity. Second, JOANA's multi-omics integration via Bayesian networks inherently accounts for missing values and reveals pathways invisible to single-layer analyses. Finally, we demonstrate high versatility across diverse experimental paradigms-from proteomics and transcriptomics to single-cell transcriptomics, mutation analysis, and transcriptomics-epigenomics data. In systematic comparisons on synthetic data as well as diverse real-world multi-modal datasets, JOANA achieves up to $\sim $20-fold reduction in reported pathways compared with existing methods while maintaining sensitivity for true biological signals. We implement JOANA in an open-source Python package, joanapy.