A semiautomated framework for integrating expert knowledge into disease marker identification.

Area of Science:

• Bioinformatics
• Genomics
• Proteomics

Background:

• High-throughput technologies generate large datasets, facilitating disease biomarker studies.
• Integrating expert knowledge into biomarker selection from complex data remains a challenge.

Purpose of the Study:

• Develop a generalizable framework for semi-automated integration of expert knowledge into data-driven biomarker selection.
• Provide an optimization metric for biomarker selection schemes.

Main Methods:

• Implemented a five-component pipeline for Identification by Signatures from Integrated Clustering (ISIC).
• Combined distance-based clustering with expert knowledge-driven functional selection for biomarker identification.

Main Results:

• Demonstrated ISIC utility on proteomics data for chronic obstructive pulmonary disease (COPD).
• Identified and validated biomarker candidates in mouse models and human cohorts.

Conclusions:

• Expert knowledge integration enhances the robustness of biomarker candidates.
• Strategies for extracting orthogonal and robust features improve biomarker discovery success.