Glutathione peroxidase 4 (GPX4) is canonically viewed as the primary suppressor of ferroptosis, yet its role in the tumor microenvironment (TME) extends far beyond antioxidant catalysis to encompass immuno-metabolic regulation. In this review, we synthesize recent advances in enzymology, immunology, and cancer metabolism to propose a “lipid peroxidation threshold” framework, wherein GPX4 sets cell-type-specific thresholds that determine susceptibility to ferroptosis across tumor cells, CD8⁺ T cells, dendritic cells (DCs), and myeloid populations. We discuss how these thresholds are dynamically adjusted by post-translational modifications, nutrient competition and intercellular feedback loops, resulting in significant spatial heterogeneity between the tumor core and the tumor invasive front. There is a current selectivity paradox in GPX4 inhibitors, as well as resistance through nuclear factor erythroid 2-related factor 2 (Nrf2) and ferroptosis suppressor protein 1 (FSP1) that restricts the efficacy of GPX4 inhibitors as monotherapy. We focus on rational combination approaches: GPX4 modulation with immune checkpoint blockade (ICB), chemotherapy, and targeting myeloid-derived suppressor cells (MDSCs); and the pressing need for predictive biomarkers and single-cell spatial profiling. We conclude that successful clinical translation requires moving beyond indiscriminate GPX4 inhibition toward precision “threshold engineering” that selectively lowers tumor lipid peroxidation thresholds while sparing immune cells.

Non-coding RNAs (ncRNAs) regulate gene expression at multiple levels in cancer, yet their functional plasticity and the slow pace of clinical translation remain incompletely explained. This review is organized around a unifying theme: ncRNAs act as integrative nodes connecting the epigenetic, transcriptional, signaling, and metabolic regulatory layers. Building on the established signal, sponge, scaffold, and guide archetypes of long non-coding RNA function, we extend this schema into a 4S+G model that adds a splice-regulatory archetype and embeds all five modes in a three-axis taxonomy spanning structural topology, subcellular localization, and molecular mode of operation, thereby capturing context-dependent functions that binary structural categories cannot accommodate. We dissect chromatin remodeling dynamics, evaluate the competing endogenous RNA hypothesis against stoichiometric evidence, and trace metabolic–epigenetic feedback circuits that stabilize malignant states. We then critically appraise the clinical record of ncRNA therapeutics, from the termination of MRX34 to the modest efficacy of subsequent candidates, and analyze the delivery, immunogenicity, and biomarker barriers that distinguish oncology from other indications for RNA drugs. Emerging solutions, including ligand-directed extrahepatic delivery, spatial transcriptomics, and deep learning-based target prediction, are assessed against these bottlenecks. The 4S+G framework links mechanistic understanding to a realistic roadmap for developing ncRNA-targeted cancer therapeutics with favorable risk–benefit profiles.