Related Experiment Video
Updated: Jun 4, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.7K
Machine Learning-Based Pathomics Model to Predict the Prognosis in Clear Cell Renal Cell Carcinoma
Xiangyun Li1, Xiaoqun Yang1, Xianwei Yang1
1Department of Pathology, Shanghai Ruijin Hospital, Shanghai Jiaotong University School of Medicine, Shanghai, China.
Technology in Cancer Research & Treatment
|December 20, 2024
Summary
This study developed a machine learning pathomics model to predict overall survival in clear cell renal cell carcinoma (ccRCC) patients. The model identified a subtype associated with worse survival, highlighting immune genes and signaling pathways for further research.
Area of Science:
- Oncology
- Computational Pathology
- Bioinformatics
Background:
- Clear cell renal cell carcinoma (ccRCC) presents a significant challenge due to its high lethality and poor survival rates.
- Pathomics, integrating computer vision and machine learning, offers a promising avenue for improving ccRCC classification, prognosis, and treatment.
- Predicting overall survival (OS) in ccRCC patients is crucial for effective clinical management.
Purpose of the Study:
- To develop and validate a novel pathomics model for predicting overall survival (OS) in clear cell renal cell carcinoma (ccRCC) patients.
- To identify distinct patient subtypes based on pathomics features and investigate their association with clinical outcomes.
- To explore the underlying biological mechanisms, including gene expression, immune infiltration, and mutational profiles, associated with ccRCC subtypes.
Main Methods:
- Utilized The Cancer Genome Atlas (TCGA) database for ccRCC patient data, including H&E-stained slides and clinical information.
- Extracted 368 pathomics features from H&E slides using PyRadiomics and constructed a model with the non-negative matrix factorization (NMF) algorithm.
- Assessed model performance using Kaplan-Meier (KM) survival curves and Cox regression; conducted differential gene expression, gene ontology (GO) enrichment, immune infiltration, and mutational analyses.
Main Results:
- A pathomics model identified two ccRCC subtypes (Cluster 1 and Cluster 2), with Cluster 2 significantly associated with worse overall survival (OS).
- Identified 76 differentially expressed genes between subtypes, enriched in extracellular matrix organization and structure.
- Observed high expression of immune-related genes (CTLA4, CD80, TIGIT) in Cluster 2 and high mutation rates (>40%) in VHL, PBRM1, and PI3K-Akt, HIF-1, MAPK pathways.
Conclusions:
- The machine learning-based pathomics model effectively predicts OS and differentiates ccRCC subtypes.
- The study highlights the prognostic significance of immune-related genes like CTLA4 and key signaling pathways (PI3K-Akt, HIF-1, MAPK).
- These findings provide valuable insights for advancing ccRCC molecular mechanisms, diagnosis, and therapeutic strategies.
