A transcriptomic signature of mechanical-memory-related genes predicts progression in prostate cancer
Year: 2026
Type: article
Abstract: Background: Prostate adenocarcinoma (PRAD) is clinically heterogeneous, and conventional markers incompletely explain variation in disease progression. Mechanical memory is the persistence of a mechanically induced cellular state after the initiating cue has been removed. This study aimed to develop and validate a mechanical memory gene (MMG)-related transcriptomic risk score and characterise its prognostic and microenvironmental associations in PRAD.Methods: We intersected differentially expressed genes from The Cancer Genome Atlas (TCGA)-PRAD cohort with a curated MMG set and defined molecular subtypes by consensus clustering. A progression-free interval (PFI) model was constructed using univariate Cox, least absolute shrinkage and selection operator Cox, and multivariable Cox analyses. The model was evaluated in GSE116918 and GSE94767. Bulk and single-cell transcriptomic analyses were used to examine genomic alterations, immune states, predicted drug response, and the cellular distribution and functions of core genes.Results: Thirty-nine MMG-related differentially expressed genes defined two molecular subtypes with distinct PFI. An eight-gene model comprising MYL9, EZH2, COL1A1, LMNB1, KIF14, POSTN, DDR2, and PARVA stratified patients by PFI and retained independent prognostic value after adjustment for clinicopathological variables. Both external cohorts showed consistent relative risk stratification. Shapley additive explanations (SHAP) identified DDR2, COL1A1, and MYL9 as leading contributors to the model-derived risk score. The high-risk group had higher tumour mutational burden, increased immune-checkpoint expression, higher tumour immune dysfunction and exclusion (TIDE) and myeloid-derived suppressor-cell scores, and a lower microsatellite instability (MSI) score. Predicted drug-response analyses identified differential candidate compounds, while cell-cell communication and in silico perturbation analyses prioritised immune, vascular, and regulatory pathways for follow-up. Single-cell analyses localised the three core genes mainly to endothelial and vascular populations and associated them with angiogenesis, matrix remodelling, hypoxic adaptation, metabolic reprogramming, and intercellular communication.Conclusions: The MMG-related transcriptomic risk score stratified PFI and delineated genomic, immune, and vascular-stromal differences in PRAD. COL1A1, DDR2, and MYL9 prioritised a collagen-related vascular-stromal programme for further investigation. The score provides an MMG-related transcriptomic framework for prognostic and microenvironmental stratification and prioritises testable hypotheses for functional validation.
Authors: Taize Sun, Yuhong Wu, Hualei Chen, Fei Wang
Institutions: Hainan General Hospital, Hainan Medical University
Language: English
Subfield: Cancer Research
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W7223788623