Regularized logistic regression with adjusted adaptive elastic net for gene selection in high dimensional cancer...

Area of Science:

• Genetics and Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• High-dimensional data analysis is crucial for cancer classification and gene selection.
• Adaptive regularized logistic regression, including the adaptive elastic net, has shown promise in this area.
• Existing methods like the adaptive elastic net have limitations, including biased gene selection and suboptimal performance with low variable correlations.

Purpose of the Study:

• To propose Adjusted adaptive regularized logistic regression (AAElastic) to overcome limitations of existing methods.
• To enhance gene selection consistency and classification performance in high-dimensional cancer data.
• To encourage grouping effects in gene selection.

Main Methods:

• Development of the Adjusted adaptive regularized logistic regression (AAElastic) model.
• Application of AAElastic to high-dimensional cancer classification datasets.
• Comparison of AAElastic with three other competitor regularization methods, including the adaptive elastic net.

Main Results:

• AAElastic demonstrated significantly higher consistency in gene selection compared to three competitor methods.
• The classification performance of AAElastic was comparable to the adaptive elastic net.
• AAElastic outperformed other regularization methods in classification accuracy.

Conclusions:

• AAElastic is a reliable adaptive regularized logistic regression method for high-dimensional cancer classification.
• The proposed method addresses limitations of the adaptive elastic net, particularly regarding gene selection bias and performance with varying correlations.
• AAElastic offers improved gene selection and classification accuracy, making it a valuable tool in cancer research.