Determining gene expression on a single pair of microarrays

Area of Science:

• Genomics
• Bioinformatics
• Statistical analysis

Background:

• Microarray experiments often face limitations in the number of replicates due to cost and sample availability.
• Analyzing Affymetrix microarray data with only one replicate per condition (N=1) presents analytical challenges with few existing solutions.

Purpose of the Study:

• To introduce and evaluate a novel algorithm, PINC (PINC is Not Cyber-T), for analyzing Affymetrix microarray experiments.
• To address the challenge of analyzing microarray data with limited replicates.

Main Methods:

• PINC utilizes a Bayesian framework, adapting the Cyber-T algorithm.
• It treats probe pairs within a probeset as independent gene expression measures.
• Assigns corrected p-values for each gene comparison.

Main Results:

• PINC-generated p-values effectively control the False Discovery Rate on Affymetrix control datasets.
• The p-values are sufficiently small to allow the use of conservative family-wise error rates (e.g., Holm's step-down method) with minimal loss of sensitivity.
• PINC demonstrates superior performance compared to existing methods for identifying differentially expressed genes in N=1 Affymetrix microarrays.

Conclusions:

• PINC significantly outperforms previous methods for analyzing Affymetrix microarrays with N=1 replicates.
• Beyond analyzing isolated pairs, PINC can also assess variability among biological replicates in biological samples.