A machine learning method based on the genetic and world competitive contests algorithms for selecting genes or...

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning in Biology

Background:

• Gene/feature selection is crucial for machine learning model development and biomarker identification.
• Existing methods often struggle with parameter tuning and suboptimal performance.
• Addressing these limitations is vital for advancing biological applications.

Purpose of the Study:

• To introduce a universal wrapper approach for gene/feature selection.
• To overcome limitations of existing methods, including parameter tuning difficulties and performance issues.
• To enhance the efficiency and effectiveness of machine learning in biological contexts.

Main Methods:

• Developed a universal wrapper approach integrating a novel optimization algorithm with the genetic algorithm (GA).
• Candidate solutions feature variable lengths, scored by a support vector machine.
• Evaluated the method on thirteen diverse biological datasets (classification and regression).

Main Results:

• The proposed method demonstrated superior performance compared to existing approaches.
• Successfully reduced the complexity associated with parameter tuning for users.
• Validated usefulness across various biological scopes, including drug discovery and cancer diagnostics.

Conclusions:

• The novel universal wrapper approach offers an effective solution for gene/feature selection.
• It simplifies the application of machine learning in biology by mitigating parameter tuning challenges.
• Enables optimization of biological applications, such as developing biomarker diagnostic kits with high accuracy and minimal gene sets.