Comparison of false discovery rate methods in identifying genes with differential expression

Area of Science:

• Genomics and Proteomics
• Statistical Bioinformatics
• High-Throughput Data Analysis

Background:

• High-throughput techniques generate thousands of hypotheses, necessitating robust methods for identifying significant results.
• Controlling false positives is critical in genomics and proteomics to ensure reliable downstream analysis.
• False Discovery Rate (FDR) controlling methods aim to manage the proportion of false positives among significant findings.

Purpose of the Study:

• To compare the performance of various FDR-controlling methods in typical microarray experiments.
• To investigate the observed sensitivity and power of different FDR methods when the true discoveries are unknown.
• To provide insights into the practical application of FDR methods in microarray data analysis.

Main Methods:

• Comparative analysis of five FDR-controlling methods by Benjamini et al., the q-value method by Storey, and the Bonferroni method.
• Evaluation based on two well-studied microarray datasets.
• Assessment of 'observed' sensitivity and 'apparent' test power.

Main Results:

• The BH95 (Benjamini-Hochberg) method exhibited the best control of FDR at the targeted level.
• In terms of apparent test power, the ranking of methods was: Step-down < Step-up: dependent < BH95 < Step-up adaptive < q-value.
• The study identified specific FDR methods that perform better under certain conditions in microarray analyses.

Conclusions:

• The BH95 method is recommended for its robust FDR control in microarray experiments.
• The findings offer guidance for researchers selecting appropriate FDR methods for high-throughput data analysis.
• Understanding the performance characteristics of different FDR methods is crucial for accurate interpretation of genomic and proteomic results.