Comprior: facilitating the implementation and automated benchmarking of prior knowledge-based feature selection...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Reproducible benchmarking is crucial for evaluating feature selection methods in gene expression data analysis.
• Existing systems lack extensibility and comprehensive assessment for prior knowledge approaches.
• Prior knowledge feature selection methods require uniform access to biological knowledge bases, which is currently unaddressed.

Purpose of the Study:

• To develop an extensible benchmarking system for feature selection approaches, focusing on prior knowledge methods.
• To provide uniform access to biological knowledge bases for feature selection.
• To enable comprehensive evaluation of feature selection methods based on performance, robustness, and biological relevance.

Main Methods:

• Development of the Comprior benchmark tool.
• Integration of extensible custom approaches and built-in standard feature selection methods.
• Implementation of uniform access to multiple biological knowledge bases.
• Creation of a customizable evaluation infrastructure for performance, robustness, runtime, and biological relevance assessment.

Main Results:

• Comprior facilitates rapid development and effortless benchmarking of feature selection approaches.
• The tool offers built-in standard methods and supports custom approaches.
• It provides uniform access to various knowledge bases and a customizable evaluation framework.
• Enables comparison of feature selection methods on classification performance, robustness, runtime, and biological relevance.

Conclusions:

• Comprior enables reproducible benchmarking, especially for prior knowledge feature selection approaches.
• The tool enhances the applicability of prior knowledge methods in gene expression analysis.
• It provides the first comprehensive assessment of prior knowledge feature selection effectiveness.