BOSO: A novel feature selection algorithm for linear regression with high-dimensional data.

Area of Science:

• Biomedical data analysis
• Machine learning
• Computational biology

Background:

• High-dimensional datasets are rapidly increasing across biomedical fields.
• Effective predictive modeling requires robust feature selection strategies.
• Existing methods may struggle with the complexity of large-scale biomedical data.

Purpose of the Study:

• To introduce a novel feature selection algorithm for linear regression.
• To evaluate the performance of the new algorithm against existing methods.
• To demonstrate the algorithm's utility in predicting cancer drug sensitivity.

Main Methods:

• Development of the Bilevel Optimization Selector Operator (BOSO) algorithm.
• Benchmarking BOSO against established feature selection techniques.
• Application of BOSO for predicting drug sensitivity, with a focus on methotrexate in cancer metabolism.

Main Results:

• BOSO demonstrated superior accuracy in feature selection for high-dimensional datasets compared to benchmark algorithms.
• The algorithm proved effective in a proof-of-concept for predicting drug sensitivity in cancer.
• Detailed analysis confirmed BOSO's capability in analyzing complex drug-response relationships.

Conclusions:

• BOSO is a highly accurate and effective feature selection tool for high-dimensional biomedical data.
• The algorithm offers a promising approach for advancing predictive modeling in precision medicine.
• BOSO has significant potential for applications in cancer research and drug development.