Utilization of two sample t-test statistics from redundant probe sets to evaluate different probe set algorithms in...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Probe set algorithms significantly influence GeneChip expression summary and downstream analysis.
• Current performance evaluation relies on spiked-in cRNAs, which may not reflect endogenous gene expression.
• There is a need for methods to assess algorithm suitability for specific datasets without external reference data.

Purpose of the Study:

• To propose and validate novel approaches for evaluating probe set algorithm performance using probe set redundancy.
• To assess data variance and result bias without relying on external reference datasets.
• To determine if probe set algorithms exhibit dataset-specific performance.

Main Methods:

• Developed three approaches analyzing redundant probe set variance and result bias using t-statistic statistics.
• Approaches include t-statistic rank order, correlation between redundant probe sets, and co-occurrence of replicate probe sets.
• Applied methods to expression summary data from MAS5.0, dChip, and RMA algorithms across three datasets.

Main Results:

• The three proposed approaches yielded similar results within datasets and agreed with prior findings.
• Analysis of redundant probe set variance and t-statistic correlation provided effective data variance assessment.
• Co-occurrence analysis of replicate probe sets enabled estimation of result bias.

Conclusions:

• Probe set redundancy is a viable feature for assessing probe set algorithm performance.
• The developed methods offer robust tools for data variance analysis and bias estimation.
• Individual probe set algorithms demonstrate dataset-specific performance, highlighting the importance of careful selection.