Identification of differential gene expression for microarray data using recursive random forest.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• DNA microarray data presents challenges due to high dimensionality and complex structures.
• Traditional random forest methods struggle with optimal gene selection due to undifferentiated genes.
• Effective gene selection is crucial for accurate analysis of high-dimensional biological data.

Purpose of the Study:

• To introduce and validate recursive random forest analysis for gene selection in DNA microarray data.
• To demonstrate the effectiveness of recursive random forest in improving classification accuracy.
• To identify biologically relevant genes associated with diseases from microarray datasets.

Main Methods:

• Recursive random forest, an enhanced random forest algorithm, iteratively removes non-influential genes.
• Gene importance is assessed, and classification performance is evaluated using the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curve.
• The method was applied to five diverse microarray datasets: colon, prostate, leukemia, breast, and skin.

Main Results:

• Recursive random forest achieved superior classification results with a reduced set of selected genes across all analyzed datasets.
• Biological validation using NCBI confirmed the relevance of selected genes for breast and skin cancer data.
• The method effectively retained genes crucial for disease association.

Conclusions:

• Recursive random forest is a powerful tool for DNA microarray data analysis and gene selection.
• The identified genes provide valuable insights for clinical diagnostics and understanding disease mechanisms.
• This approach enhances the interpretability and utility of high-dimensional genomic data.