Related Experiment Video
Updated: May 22, 2025
MicroRNA Based Liquid Biopsy: The Experience of the Plasma miRNA Signature Classifier MSC for Lung Cancer Screening
Published on: October 26, 2017
A machine learning-based prognostic signature utilizing MSC proteomics for predicting bladder cancer prognosis and
Xinyu Zhang1, Pan Li1, Luhua Ji1
1Institute of Urology, Lanzhou University Second Hospital, Key Laboratory of Gansu Province for Urological Diseases, Gansu Urological Clinical Center, Lanzhou, China.
Background:
Mesenchymal stem cells (MSCs), due to their tumor-targeting homing properties, are present in the tumor microenvironment (TME) and influence the biological behaviors of tumors. The purpose of this paper is to establish a signature based on the MSC secretome to predict the prognosis and treatment of bladder cancer (BLCA).
Methods:
The presence of MSCs in BLCA was validated through flow cytometry and multiplex fluorescence immunohistochemistry (mFIHC), and the relationships between MSCs and clinical characteristics were explored. Unsupervised clustering analysis was performed on BLCA according to the differential proteins detected in MSC-conditioned medium (MSCCM) using a cytokine array. Using the TCGA-BLCA, GSE32548, and GSE32894 datasets as background data, a risk signature was constructed according to the differential proteins in MSCCM through machine learning. For the risk groups with high and low prognoses, we calculated Kaplan-Meier (K-M) curves. Additionally, we explored the relationships between the signature and the tumor immune landscape, response to immunotherapy, and chemotherapy drugs.
Results:
Both flow cytometry and mFIHC confirmed the presence of MSCs in bladder tumors, and clinical samples revealed correlations between MSCs and the pathological grade, T stage, and Ki67 in BLCA. Based on differential proteins and unsupervised clustering analysis, BLCA patients were divided into two groups, and significant differences were found between these groups in terms of TME, immune response, and clinical treatments. Using machine learning, a signature was constructed with the combination algorithm Stepcox (both) + plsRcox, revealing significant survival differences between the high- and low-risk MSC groups. Regression analyses, along with ROC curves, further demonstrated that risk score independently predict the prognosis of patients with high predictive performance. Moreover, there were notable differences between the high- and low-risk groups in terms of the TME scores, immune infiltration, and immune checkpoints. For BLCA immunotherapy, the low-risk group suggested better efficacy, while conventional chemotherapy drugs such as gemcitabine and cisplatin might be less effective in the low-risk group.
Conclusion:
The signature based on MSC secreted protein profiles could effectively predict the prognosis of BLCA and provided valuable guidance for treatment and drug resistance.
