Robust Significance Analysis of Microarrays by Minimum β-Divergence Method

Area of Science:

• Genomics
• Statistical Bioinformatics
• Gene Expression Analysis

Background:

• Identifying differentially expressed (DE) genes is crucial for biomarker discovery.
• Significance Analysis of Microarrays (SAM) is widely used but sensitive to outliers, reducing its power.
• Existing methods struggle with accuracy when gene expression data contains outlying values.

Purpose of the Study:

• To develop a robust statistical approach for identifying DE genes.
• To enhance the Significance Analysis of Microarrays (SAM) method by incorporating minimum β-divergence estimators.
• To evaluate the performance of the proposed robust SAM method against other popular statistical techniques.

Main Methods:

• The proposed method utilizes minimum β-divergence estimators to replace maximum likelihood estimators in the SAM framework.
• Performance was assessed using both simulated and real gene expression datasets.
• Comparative analysis included ANOVA, SAM, LIMMA, KW, EBarrays, GaGa, and BRIDGE.

Main Results:

• All tested methods performed comparably in large-sample cases without outliers.
• In small-sample cases without outliers, SAM, LIMMA, and the proposed method showed superior and similar performance.
• The proposed robust method demonstrated superior performance across both small and large sample sizes in the presence of outliers.

Conclusions:

• The proposed robust statistical method effectively addresses the limitations of traditional SAM in the presence of gene expression outliers.
• This enhanced approach offers improved power and reliability for identifying differentially expressed genes, particularly in datasets with noisy or outlying data.
• The findings suggest the robust SAM method as a valuable tool for biomarker discovery in genomics research.