Multiobjective Patient Stratification Using Evolutionary Multiobjective Optimization.

Area of Science:

• Computational Biology
• Bioinformatics
• Medical Informatics

Background:

• Patient stratification is crucial for personalized medicine but faces challenges like noise, high dimensionality, and poor interpretability in existing clustering methods.
• Current algorithms struggle to effectively balance feature relevance and redundancy, hindering biologically meaningful patient subgroup identification.

Purpose of the Study:

• To propose and formulate a multiobjective framework using evolutionary multiobjective optimization for enhanced patient stratification.
• To address limitations of existing methods by balancing feature relevance and redundancy for improved accuracy and interpretability.

Main Methods:

• Developed a multiobjective framework based on evolutionary multiobjective optimization.
• Benchmarked algorithms across 55 synthetic datasets (human transcription regulation network model) and 35 real cancer gene expression datasets.
• Conducted time complexity, convergence, and parameter analyses for robustness.

Main Results:

• The proposed algorithms demonstrated superior performance compared to state-of-the-art methods.
• Experimental results validated the effectiveness and robustness of the evolutionary multiobjective optimization framework.
• t-Distributed Stochastic Neighbor Embedding (t-SNE) was used for visualizing high-dimensional data.

Conclusions:

• The novel multiobjective framework offers a robust and effective solution for patient stratification in personalized medicine.
• The approach successfully balances feature relevance and redundancy, leading to improved accuracy and interpretability.
• This work advances the field by providing a more sophisticated tool for analyzing complex biological data for clinical applications.