液-液相分离相关基因特征刻画了胃癌的预后亚型及治疗敏感性

Gastric cancer (GC) is a heterogeneous malignancy with variable outcomes. Liquid-liquid phase separation (LLPS) has emerged as a regulator of cancer-related processes. The study aims to identify GC subgroups associated with LLPS and to establish a prognostic model for patient outcomes.

The Cancer Genome Atlas (TCGA) GC transcriptomic and clinical data were analyzed. LLPS-related genes were obtained from the Phase Separation Database. Differential expression analysis and unsupervised consensus clustering were performed to define molecular subtypes. A prognostic RiskScore was constructed using least absolute shrinkage and selection operator (LASSO) Cox regression and validated in independent Gene Expression Omnibus cohorts (GSE84437, GSE66229, GSE28541). Intrinsically disordered regions (IDRs) were predicted using the Intrinsically Unstructured Protein Predictor. Model performance and clinical utility were assessed using time-dependent receiver operating characteristic (ROC) curve analysis and decision curve analysis (DCA). Immune cell infiltration, immunotherapy responsiveness, functional pathway enrichment, and drug sensitivity were analyzed using established computational frameworks.

A three-gene LLPS-related prognostic signature ( ZFYVE27 , GNG11 , and DOK7 ) was used to derive a RiskScore that defined high- and low-risk GC groups with significantly different overall survival across multiple cohorts. Intrinsic disorder analysis revealed that all three proteins contain substantial IDRs, supporting their LLPS-related properties. Time-dependent ROC and concordance index analyses indicated moderate discriminative performance. A nomogram integrating RiskScore and clinical variables demonstrated improved predictive accuracy and calibration. DCA showed that the combined model provided a higher net benefit than the clinical model alone. The risk groups exhibited distinct immune cell infiltration profiles and differences in predicted immunotherapy responsiveness. Gene set enrichment analysis demonstrated enrichment of cell cycle, oncogenic, and inflammatory pathways in the high-risk group, including G2M checkpoint, MYC targets, and IL6/JAK/STAT3 signaling. Drug sensitivity prediction suggested differential therapeutic vulnerabilities, with high-risk patients showing increased predicted sensitivity to JAK inhibitors, dasatinib, and nutlin-3a.

An LLPS-related three-gene RiskScore identifies GC subgroups with different prognosis, immune features, and therapeutic sensitivity.