Related Experiment Video
Updated: Aug 18, 2026
Microarray-based Identification of Individual HERV Loci Expression: Application to Biomarker Discovery in Prostate Cancer
Published on: November 2, 2013
Development of a prostate cancer biochemical recurrence risk signature using machine learning and motor
Weixing Wang1, Guohai Xie2, Zhonggao Wu1
1Department of Urology, Ningbo Haishu People's hospital, Ningbo, Zhejiang, China.
Background:
Motor proteins play significant roles in cancer progression, but their involvement in biochemical recurrence (BCR) of prostate cancer remains unclear. The objective of the study is to develop a prognostic indicator for BCR using machine learning (ML) and motor protein-related genes (MPRGs).
Methods:
The prognosis relevance of the MPRGs in prostate cancer was analyzed by univariate Cox regression. Feature selection and model construction were performed using combinations of multiple machine learning algorithms. Model performance was assessed using receiver operating characteristic curve and C-index. Patients were stratified into high- and low-risk groups based on the risk signature, and comparisons of BCR incidence, gene expression profiles, immune cell infiltration patterns, and drug sensitivity were conducted between these groups. The gene expression of MPRGs were validated in vitro.
Results:
Among 120 MPRGs, 17 were differentially expressed, of which 8 were significantly associated with BCR. A novel risk scoring system using a StepCox[forward] + Ridge model based on these 8 MPRGs effectively stratified patients into two different risk groups, and patients with high riskscores had significantly higher BCR rates than those with lower riskscores. Enrichment analysis revealed upregulation of inflammation response, EMT, hypoxia, and estrogen response pathways in the high-risk category, while mitotic spindle, G2M checkpoint, and E2F targets were downregulated. The MPRG-derived risk score correlated positively with M2 macrophage infiltration and ngatively correlated with CD4 T cells and mast cells, and the high-risk category showed higher sensitivity to drugs like cisplatin and bicalutamide. The final nomogram based on MPRGs-derived signature and T stage provided an excellent tool for predicting BCR. In vitro experiments further validated that the expression trends of MPRGs in the risk signature were consistent with the bioinformatics analysis results.
Conclusion:
This study developed a novel MPRG-derived risk signature that effectively predicts BCR in prostate cancer, offering valuable insights for clinical management and personalized treatment strategies.
