Speeding Up the Discovery of Optimal Feature Combinations for Omics Data Based on Pseudo-Kernel Function

Area of Science:

• Computational biology
• Bioinformatics
• Machine learning in healthcare

Background:

• Accurate disease classification and prediction rely on identifying meaningful feature combinations.
• Existing methods face computational challenges with large omics datasets due to high feature numbers.

Purpose of the Study:

• To develop an efficient algorithm for feature combination analysis in disease classification.
• To reduce computational cost while preserving classification performance.

Main Methods:

• Proposed PKF-k-TSP, a novel omics data analysis method using pseudo kernel functions.
• Explores linear and nonlinear feature combinations and evaluates feature interactions.
• Selects top-scoring feature pairs to build an ensemble classifier, mapping features to a high-dimensional space.

Main Results:

• PKF-k-TSP demonstrated superior classification performance and significantly improved computational efficiency (72.43% reduction in running time vs. KF-k-TSP).
• Identified feature pairs correlate with physiological/pathological changes, offering disease mechanism insights.
• Effective in cross-cancer pathway interaction analysis, revealing conserved and tissue-specific signaling networks.

Conclusions:

• PKF-k-TSP enables rapid and efficient feature mining for large-scale omics data.
• The method is suitable for disease diagnosis, prognosis, and understanding complex biological pathways.