Related Experiment Video
Updated: Jan 18, 2026
08:50
Predictive Immune Modeling of Solid Tumors
Published on: February 25, 2020
7.5K
Interpretable Machine Learning for Proteomics-Based Subtyping and Tumor Mutational Burden Prediction in Endometrial
Thi-My-Trang Luong1,2,3, Xuan Lam Bui1,2, Chii-Ruey Tzeng3
1International Master/Ph.D. Program in Medicine, College of Medicine, Taipei Medical University, Taipei, Taiwan.
Proteomics. Clinical Applications
|September 8, 2025
Summary
This study developed a machine learning model using proteomics data to accurately classify endometrial carcinoma (EC) molecular subtypes and predict tumor mutational burden (TMB). The model identified key protein biomarkers, paving the way for precision medicine in EC.
Area of Science:
- Proteomics
- Machine Learning
- Oncology
Background:
- Endometrial carcinoma (EC) exhibits significant molecular heterogeneity, impacting prognosis and treatment.
- Accurate classification of EC molecular subtypes and tumor mutational burden (TMB) is vital for personalized therapy.
- Integrating proteomics with machine learning (ML) offers a path for precise EC classification and biomarker discovery.
Purpose of the Study:
- To develop and validate a machine learning pipeline using proteomics data for classifying EC molecular subtypes.
- To accurately predict tumor mutational burden (TMB) in EC patients.
- To identify novel protein biomarkers for EC through ML interpretability.
Main Methods:
- Utilized proteomic data from 95 EC patients (CPTAC).
- Developed an ML pipeline involving feature selection (Lasso), classification (logistic regression), and interpretability (SHAP, LIME).
- Applied SMOTE for class imbalance and evaluated models using accuracy, AUC, precision, recall, and F1-score.
Main Results:
- Reduced 11,000 proteomic features to eight key proteins.
- Achieved high accuracy in classifying EC molecular subtypes (82.8%) and predicting TMB (89.7%).
- Identified validated (MLH1, PMS2, STAT1) and novel (MTHFD2, MAST4, etc.) protein biomarkers.
Conclusions:
- A proteomics-driven ML approach provides accurate and interpretable EC classification and TMB prediction.
- The identified biomarkers offer biological insights and support the development of non-invasive diagnostics.
- This strategy lays the groundwork for advancing personalized treatments and precision medicine in EC.
