ReduXis: A Comprehensive Framework for Robust Event-Based Modeling and Profiling of High-Dimensional Biomedical Data.

Area of Science:

• Computational Biology
• Biomedical Informatics
• Machine Learning in Medicine

Background:

• Event-based models (EBMs) are valuable for analyzing biomarker changes in progressive diseases.
• Challenges include data quality issues, high dimensionality, and limited interpretability of EBMs.
• Existing methods require significant manual effort for data preparation and feature selection.

Purpose of the Study:

• To introduce ReduXis, a streamlined pipeline designed to enhance the application of EBMs.
• To address data quality, high dimensionality, and interpretability challenges in EBMs.
• To facilitate the analysis of biomarker data for disease progression.

Main Methods:

• Automated data readiness assessment upon dataset upload, including format verification, metadata completeness, and measurement compatibility checks.
• Ensemble voting-based feature selection using gradient boosting, logistic regression, and random forest classifiers to identify robust biomarker subsets and prevent overfitting.
• Generation of interpretable outputs such as subject-level staging, subtype assignments, comparative biomarker profiles, and classification performance visualizations.

Main Results:

• ReduXis successfully automates data quality checks and provides actionable feedback.
• The ensemble feature selection effectively identifies relevant biomarkers in high-dimensional data.
• The pipeline generates clear, interpretable visualizations and assignments for downstream analysis.
• Validation across Alzheimer's disease, transitional cell carcinoma, and colorectal adenocarcinoma cohorts demonstrates ReduXis's versatility.

Conclusions:

• ReduXis offers a robust and interpretable solution for applying event-based models to complex biomarker data.
• The pipeline enhances transparency and facilitates downstream analysis in disease progression studies.
• ReduXis has the potential to improve the utility of EBMs in clinical and research settings.