A tipping point signals the shift from a stable phase to a malignant state in cervical cancer progression. Identifying this transition is essential for early intervention, yet traditional biomarkers based on differential expression or undirected network measures often fail to capture it. Here, we introduce Causality-Directed Network Flow Entropy (CDNFE), a framework that quantifies entropy dynamics in directed regulatory networks to detect disease tipping points with superior sensitivity over expression-based methods. Applied to cervical cancer data which were derived from samples of different clinical stages collected from Luohe Central Hospital (single-cell, bulk transcriptome, simulations, and spatial transcriptomics), CDNFE consistently identified a precancerous tipping point and uncovered a biomarker module enriched for non-differentially expressed “dark genes.” Within this module, FNDC3B and NECTIN4 emerged as central hubs, validated across multiple levels: (1) functionally essential in cervical cancer cell lines, (2) structurally important as driver nodes within regulatory networks, (3) mechanistically suggesting a potential link to PI3K/AKT signaling driven by epithelial–mesenchymal transition, and (4) spatially enriched in tumor regions in independent spatial transcriptomics sections. These findings establish CDNFE as a robust systems-level framework for tipping point detection and highlight FNDC3B and NECTIN4 as representative dark gene regulators of cervical cancer progression.
Qiao et al. (2026) studied this question.