Identifying dynamical persistent biomarker structures for rare events using modern integrative machine learning...

Area of Science:

• Computational Biology
• Biotechnology
• Data Science

Background:

• Omics and computational advancements accelerate biological discoveries.
• Flow cytometry offers deep insights into cellular processes for health and disease research.
• Cytometry data presents computational challenges: high dimensionality, large size, nonlinearity, diverse patterns, and class imbalance.

Purpose of the Study:

• To address challenges in analyzing high-dimensional cytometry data.
• To develop a method for simultaneous rare cell population prediction and feature selection.
• To leverage a 'wisdom of the crowd' approach integrating multiple machine learning algorithms.

Main Methods:

• Integration of a pool of modern machine learning (ML) algorithms.
• Utilizing a 'wisdom of the crowd' strategy for enhanced predictive power.
• Employing entropy and rank-based normalization techniques to integrate diverse ML models.

Main Results:

• The approach successfully detects diverse patterns across model features.
• Identifies a dynamic biomarker structure classifying features into persistent, unselected, and fluctuating groups.
• Demonstrates effective simultaneous rare cell prediction and feature selection.

Conclusions:

• The proposed method offers a robust solution for complex cytometry data analysis.
• The dynamic biomarker identification aids in understanding biomarker roles in rare cell prediction.
• This approach can significantly advance studies on disease progression and biomarker discovery.