On the relevance of technical variation due to building pools in microarray experiments

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Pooled samples are common in gene expression experiments, blending RNA from multiple individuals.
• Technical limitations can cause unequal representation of individuals in pooled samples, introducing a specific error.
• This technical error, modeled as a variance component, can impact experimental results.

Purpose of the Study:

• To assess the practical relevance of technical error arising from imperfect RNA blending in pooled samples.
• To evaluate the significance and magnitude of the variance component associated with blending disproportionality.
• To analyze the impact of this error on gene expression data across different species.

Main Methods:

• Applied previously published theory for variable pool sizes to four microarray gene expression datasets.
• Analyzed mouse, honey bee, rat, and human data to quantify the blending error variance component.
• Utilized simulations to assess the impact of neglecting blending error on statistical tests and false discovery rates.

Main Results:

• Significant blending error variance was detected in 23% of mouse transcripts and 49% of honey bee transcripts.
• Minimal or no significant blending error was observed in rat and human data, respectively.
• Neglecting blending error led to an increased number of differentially expressed transcripts and overly-optimistic (anti-conservative) significance tests.

Conclusions:

• Imperfect RNA blending is a source of technical variation impacting experimental design and data analysis in gene expression studies.
• Adverse effects include misidentification of differentially expressed transcripts and inflated significance.
• Established theory and models for data analysis can effectively mitigate these adverse effects.