From hybridization theory to microarray data analysis: performance evaluation.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Affymetrix GeneChips require robust preprocessing for accurate analysis.
• Traditional methods often rely on statistical analysis of fluorescence intensities.
• A novel algorithm, AffyILM, uses hybridization thermodynamics and an extended Langmuir isotherm model.

Purpose of the Study:

• To compare the performance of the AffyILM algorithm against traditional preprocessing methods.
• To evaluate these methods on both older and newer generations of GeneChips.
• To assess the effectiveness of AffyILM in differential expression analysis.

Main Methods:

• Utilized Affymetrix tissue mixture and Latin Square datasets.
• Employed correlation analysis to evaluate probe-level concentrations.
• Quantified algorithm performance using the Latin Square HG-U133A dataset and spiked genes.
• Compared AffyILM(medianpolish) with established intensity-based preprocessing strategies.

Main Results:

• Median-polish algorithm effectively summarizes AffyILM concentrations at the probe-level.
• AffyILM(medianpolish) demonstrated correlation results comparable to top intensity-based methods.
• Performance analysis on the Latin Square dataset showed AffyILM(medianpolish) is a top-performing strategy.

Conclusions:

• AffyILM(medianpolish) is a leading strategy for converting probe intensities to target concentrations before statistical analysis.
• This method, grounded in hybridization theory, yields identically distributed probe-level target concentration estimates.
• Future research should explore probe-level multivariate analysis of concentrations.