Integrative single-cell and spatial transcriptomics with explainable AI reveal lethal prognostic axis in prostate...

Area of Science:

• Oncology
• Genomics
• Computational Biology

Background:

• Prostate cancer (PCa) exhibits significant clinical heterogeneity.
• Accurate prognostic models are crucial for effective PCa management.

Purpose of the Study:

• To define a lethal tumor axis in PCa.
• To develop an interpretable prognostic model using explainable machine learning.

Main Methods:

• Integration of single-cell and spatial transcriptomics.
• Analysis of 141,986 single cells across PCa stages.
• Benchmarking 101 machine learning pipelines.
• Utilizing SHAP for feature interpretation.

Main Results:

• Identification of a malignant C4 epithelial subpopulation with stemness potential.
• Discovery of immune-suppressive niches within the tumor microenvironment.
• Development of a C4-based risk score predicting recurrence-free survival.
• Highlighting PCSK1N as a key driver of progression and therapeutic resistance.

Conclusions:

• The study defines a C4-centered lethal tumor axis in PCa.
• An explainable prognostic framework is established for PCa stratification.
• Targeting PCSK1N shows potential for enhancing AR-targeted therapy efficacy.