Gene set analyses for interpreting microarray experiments on prokaryotic organisms.

Area of Science:

• Genomics
• Bioinformatics
• Statistical Analysis

Background:

• DNA microarrays generate extensive gene expression data, but interpretation remains challenging.
• Gene set analysis methods are emerging to interpret transcriptional levels using biologically defined gene sets.
• Fisher's exact test (FET) is a limited but plausible option for gene set analysis with few replicates, common in prokaryotic studies.

Purpose of the Study:

• To extend existing gene set analysis methods for experiments with few replicates.
• To compare the performance of these extended methods against each other and Fisher's exact test.
• To identify a robust and accurate gene set analysis method for low-replicate experiments.

Main Methods:

• Five gene set analysis methods were extended for low-replicate experiments.
• Simulated and real experimental data were used for comparative analysis.
• Performance was evaluated based on statistical power, robustness, and type I error rate.

Main Results:

• MAXMEAN-NR maintained the nominal false positive rate and showed good power and robustness for gene sets of size 10 or more.
• ABSSUM-NR and SUM-NR were powerful for smaller gene sets (less than 10).
• MAXMEAN-NR identified biologically relevant sets missed by other methods, including FET, and outperformed GSEA-NR.

Conclusions:

• MAXMEAN-NR is a recommended gene set analysis method for low-replicate experiments, particularly in prokaryotic genomics.
• It provides increased robustness and biological relevance compared to FET and other methods.
• The method effectively maintains the nominal type I error rate, ensuring reliable results.