Estimation of the proportion of true null hypotheses under sparse dependence: Adaptive FDR controlling in microarray...

Area of Science:

• Bioinformatics
• Statistical Genetics
• Computational Biology

Background:

• Accurate estimation of non-differentially expressed genes is crucial for multiple testing procedures in microarray analysis.
• Existing methods often assume gene expression independence, which is unrealistic given sparse dependence structures in biological data.
• Developing methods to account for sparse dependence is essential for robust gene expression analysis.

Purpose of the Study:

• To propose a novel clustering-based method to estimate the proportion of true null hypotheses in gene expression data.
• To integrate sparse dependence structures into existing estimators for improved accuracy.
• To enhance the power of adaptive multiple testing procedures in microarray data analysis.

Main Methods:

• A clustering approach was developed using a high-dimensional correlation structure under a sparse assumption as a dissimilarity matrix.
• This method was applied to three existing estimators for the proportion of true null hypotheses.
• The proposed method was evaluated through extensive simulations and applied to a colorectal cancer gene expression dataset.

Main Results:

• The proposed clustering method significantly improves existing estimators, making them less conservative.
• The adaptive Benjamini-Hochberg algorithm demonstrated increased power when using the proposed method.
• Application to a colorectal cancer dataset revealed a greater number of identified differentially expressed genes.

Conclusions:

• The novel clustering method effectively accommodates sparse dependence in gene expression data.
• This approach enhances the performance of statistical inference in microarray studies.
• The method offers a valuable tool for identifying differentially expressed genes in complex biological datasets, such as cancer research.