Single Cell and Machine Learning-Based Analysis of Lysosome Mediated Cell Death Reveals Novel Prognostic Biomarkers and Therapeutic Targets in Cervical Cancer.

Background: Cervical cancer is among the most prevalent gynecological malignancies globally, and its inherent molecular heterogeneity remains a persistent obstacle to uniform treatment efficacy and patient survival. Lysosome-mediated cell death (LDCD) has recently emerged as a mechanistically distinct form of regulated cell death with growing relevance in tumor biology, influencing both oncogenic progression and responsiveness to therapy. Nevertheless, its functional significance in cervical cancer and its interactions with the tumor microenvironment (TME) at single-cell resolution remain insufficiently characterized. Methods: We conducted an integrative analysis combining single-cell RNA sequencing (scRNA-seq) with 15 machine learning algorithms to examine LDCD-related gene expression patterns in cervical cancer. Multicohort transcriptomic data were obtained from TCGA and GEO databases, and scRNA-seq data were retrieved from the GSE138080 dataset. Core hub genes were identified through differential expression analysis, protein-protein interaction network construction, and consensus clustering. Prognostic models were built using multiple machine learning frameworks and evaluated by concordance index. Immune cell infiltration was quantified via CIBERSORT, and single-cell-level validation was performed with the Seurat pipeline. Results: Six LDCD-associated hub genes-CTSV, FER, GGA2, LAMP3, STAB1, and STAB2-were identified as significant prognostic determinants. Univariate Cox regression demonstrated that CTSV ( Conclusions: Our findings establish LDCD as a functionally important mechanism in cervical cancer at single-cell resolution and demonstrate the translational utility of integrating scRNA-seq with machine learning for the discovery of novel prognostic biomarkers and therapeutic targets.