Statistical analysis of an RNA titration series evaluates microarray precision and sensitivity on a whole-array basis

Area of Science:

• Genomics
• Bioinformatics
• Gene Expression Analysis

Background:

• Concerns exist regarding the accuracy and cross-platform agreement of microarray technologies.
• Existing methods lack unambiguous evaluation of whole-array precision and sensitivity.

Purpose of the Study:

• To develop and validate a methodology for assessing the precision and sensitivity of whole-genome gene expression technologies, specifically microarrays.
• To compare the performance of different microarray platforms and data extraction algorithms.

Main Methods:

• Development of an easy-to-construct RNA sample titration series.
• Application of statistical analysis using non-linear regression for whole-array evaluation.
• Comparison of four microarray platforms and three Affymetrix data extraction algorithms.

Main Results:

• All four microarray platforms demonstrated satisfactory precision.
• Commercial platforms showed superiority in resolving differential gene expression at lower levels.
• Cross-platform variability was found to be primarily systematic, not random measurement error.
• A commonly used proprietary algorithm performed poorly compared to alternatives.

Conclusions:

• The developed methodology provides a robust evaluation of microarray precision and sensitivity.
• While platforms show good precision, probe annotation needs improvement for enhanced accuracy.
• Cross-platform comparisons are less accurate than within-platform analyses for predicting clinical outcomes.