Related Experiment Video
Updated: Jun 27, 2025
Digital Spatial Profiling for Characterization of the Microenvironment in Adult-Type Diffusely Infiltrating Glioma
Published on: September 13, 2022
ecGBMsub: an integrative stacking ensemble model framework based on eccDNA molecular profiling for improving IDH
Zesheng Li1, Cheng Wei1, Zhenyu Zhang2
1Tianjin Neurological Institute, Key Laboratory of Post-Neuro Injury, Neuro-Repair and Regeneration in Central Nervous System, Ministry of Education and Tianjin City, Tianjin Medical University General Hospital, Tianjin, China.
This study introduces ecGBMsub, a novel stacking ensemble model for accurately classifying IDH wild-type glioblastoma (GBM) subtypes using extrachromosomal circular DNA (eccDNA) profiling. The developed XGBoost.Enet-stacking-Enet model enhances clinical diagnosis and treatment guidance.
Area of Science:
- Neuro-oncology
- Computational Biology
- Machine Learning
Background:
- IDH wild-type glioblastoma (GBM) subtypes exhibit distinct molecular profiles and prognoses, necessitating precise classification for effective clinical management.
- Current machine learning approaches for GBM subtype prediction show promise but lack the robustness and accuracy required for widespread clinical adoption.
- Ensemble learning strategies offer potential for improved predictive performance by integrating multiple models.
Purpose of the Study:
- To develop and validate a novel integrative stacking ensemble model framework (ecGBMsub) for enhanced classification of IDH wild-type GBM molecular subtypes.
- To identify the optimal stacking ensemble model with high accuracy for clinical application.
- To provide an accessible web tool for utilizing the developed classification model.
Main Methods:
- Extrachromosomal circular DNA (eccDNA) molecular profiling was employed for GBM subtype classification.
- Nine single machine learning models were trained and optimized.
- A stacking ensemble approach was implemented, selecting the top five models as base learners and training nine meta-models on their predictions, generating 234 stacked ensemble models.
- Comprehensive evaluation and comparison of all developed models were performed.
Main Results:
- The ecGBMsub framework successfully integrated multiple machine learning models to improve GBM subtype classification accuracy.
- The stacking ensemble model "XGBoost.Enet-stacking-Enet" demonstrated superior performance and was selected as the optimal model within the framework.
- A user-friendly web tool was developed and is available for public access to facilitate the application of the optimal model.
Conclusions:
- The developed ecGBMsub framework, particularly the "XGBoost.Enet-stacking-Enet" model, offers a robust and accurate method for classifying IDH wild-type GBM molecular subtypes.
- This advancement has significant implications for guiding clinical diagnosis, treatment strategies, and patient stratification in GBM management.
- The accessible web tool promotes the translation of this computational tool into clinical practice.
