ieGENES: A machine learning method for selecting differentially expressed genes in cancer studies

Area of Science:

• Bioinformatics
• Computational Biology
• Cancer Research

Background:

• Accurate gene selection is essential for effective cancer classification from microarray data.
• High-dimensional microarray data presents challenges like multi-collinearity, impacting classification model stability.
• Existing gene selection methods may not optimally identify relevant genes for cancer diagnosis.

Purpose of the Study:

• To develop a novel wrapper method, ieGENES, for improved gene selection in cancer microarray data.
• To enhance cancer classification accuracy and identify differentially expressed genes (DEGs) for biomarker discovery.
• To address multi-collinearity in high-dimensional data using a parsimonious kernel machine regularization (PKMR) model.

Main Methods:

• Developed the ieGENES wrapper algorithm for optimal gene identification and redundant gene elimination.
• Utilized a parsimonious kernel machine regularization (PKMR) model with ridge regularization to handle multi-collinearity.
• Employed leave-one-out cross-validation for evaluating gene selection accuracy and updating model parameters.

Main Results:

• ieGENES outperformed existing wrapper methods in gene selection accuracy on 5 out of 6 cancer microarray datasets.
• Statistically significant improvements (p<0.001) in classification accuracy were observed.
• Achieved 96.21% accuracy with 167 DEGs for the Colon cancer dataset.

Conclusions:

• The ieGENES technique demonstrates superior performance in identifying DEGs for cancer diagnosis.
• Offers a promising tool for analyzing microarray data and discovering biomarkers in cancer research.
• Enhances the potential for accurate and reliable cancer classification.