Estimating the proportion of equivalently expressed genes in microarray data based on transformed test statistics

Area of Science:

• Bioinformatics
• Statistical Genetics
• Genomics

Background:

• False discovery rate (FDR) is crucial for multiple hypothesis testing in microarray data analysis.
• Accurate estimation of the proportion of equivalently expressed genes (pi(0)) is essential for reliable FDR control.
• Existing pi(0) estimators (BUM, SPLOSH, QVALUE, LBE) often overestimate true pi(0) when gene expression groups are not well-separated.

Purpose of the Study:

• To develop a novel and more accurate estimator for pi(0) in microarray data analysis.
• To address the overestimation issue of existing pi(0) estimators.
• To improve the precision and reduce conservativeness in FDR control.

Main Methods:

• Introduced a novel transformation of test statistics to achieve symmetry around zero.
• Developed a new pi(0) estimator based on the symmetry assumption of transformed test statistics.
• Validated the proposed method through real data application and simulation studies.

Main Results:

• The proposed pi(0) estimator demonstrates less conservative estimates compared to BUM, SPLOSH, QVALUE, and LBE.
• Simulation results indicate that the novel estimator consistently yields the lowest mean squared error among the compared methods.
• The method effectively handles cases where differentially and equivalently expressed genes are not well-separated.

Conclusions:

• The novel pi(0) estimation method provides a more accurate and less conservative approach for FDR control in gene expression studies.
• This method offers improved statistical power and reliability in identifying significant gene expression changes.
• The proposed technique represents a significant advancement in the statistical analysis of high-throughput genomic data.