Identifying interactions in omics data for clinical biomarker discovery using symbolic regression

Area of Science:

• Biomedical Informatics
• Computational Biology
• Machine Learning

Background:

• Identifying predictive biomarker signatures from omics data is crucial for clinical applications.
• Machine learning (ML) methods have advanced predictive performance but often lack interpretability.
• Complex ML models hinder clinical adoption due to their opaque nature.

Purpose of the Study:

• To apply a novel symbolic-regression-based algorithm, the QLattice, to clinical omics datasets.
• To generate parsimonious, high-performing models for disease outcome prediction and mechanism discovery.
• To demonstrate the value of feature selection in omics-based ML.

Main Methods:

• Application of the QLattice, a symbolic-regression algorithm, to clinical omics data.
• Utilizing the feyn Python package for QLattice implementation.
• Feature selection emphasizing maximal relevance and minimal redundancy.

Main Results:

• The QLattice generated parsimonious and high-performing predictive models.
• Identified biomarker signatures are interpretable and aid in understanding disease mechanisms.
• Demonstrated the effectiveness of the QLattice in omics-based machine learning.

Conclusions:

• The QLattice offers a powerful tool for developing interpretable biomarker signatures from omics data.
• Its simplicity and predictive power make it suitable for clinical decision-making and patient stratification.
• The QLattice facilitates the translation of omics data into actionable clinical insights.