Comparison of microarray preprocessing methods

Area of Science:

• Bioinformatics
• Microarray Data Analysis
• Gene Expression Profiling

Background:

• Microarray data preprocessing is vital for accurate analysis.
• No single preprocessing method is universally optimal.
• Limited datasets necessitate guidelines for quality and robustness.

Purpose of the Study:

• To compare the performance of four popular microarray preprocessing methods: MAS5, Li & Wong pmonly (LWPM), Li & Wong subtractMM (LWMM), and Robust Multichip Average (RMA).
• To provide guidelines for selecting robust preprocessing methods under laboratory constraints.

Main Methods:

• Analysis of laboratory-generated microarray data from deep lamellar keratoplasty (DLKP) cells treated with Bromodeoxyuridine (BrdU).
• Assessment of dispersion across replicates to evaluate variance reduction.
• Analysis of false discovery rate (FDR) and complementary q-value analysis.

Main Results:

• LWPM and RMA methods demonstrated superior reduction in data variability.
• LWMM method performed best in false positive analysis (parametric and nonparametric).
• LWMM showed improved overall discrimination, despite slightly less effective variance reduction compared to LWPM and RMA.

Conclusions:

• LWMM is a strong candidate for microarray data preprocessing, particularly when discrimination is a priority.
• The choice of preprocessing method impacts both variance reduction and the accuracy of differential expression analysis.
• Findings offer guidance for researchers working with limited microarray datasets.