Mixture-model based estimation of gene expression variance from public database improves identification of...

Area of Science:

• Bioinformatics
• Genomics
• Statistical Genetics

Background:

• Small sample sizes in microarray experiments hinder accurate identification of differentially expressed genes (DEGs).
• Public microarray databases offer a valuable resource for improving DEG identification efficiency.
• Conventional statistical methods face challenges with limited sample data.

Purpose of the Study:

• To develop a novel method for identifying DEGs in small sample-sized microarray datasets.
• To utilize information from public microarray databases for enhanced DEG analysis.
• To improve the accuracy of DEG detection by incorporating prior variance estimation.

Main Methods:

• Applied Gaussian mixture models to public microarray data to extract gene expression distributions.
• Estimated prior variance for Baldi's Bayesian framework using pooled variances from mixture modeling.
• Identified DEGs in small sample datasets using the enhanced Bayesian framework.

Main Results:

• The proposed method demonstrated superior performance in detecting gold-standard DEGs compared to benchmark methods.
• Accurate prior variance estimation from public databases significantly improved DEG identification in small datasets.
• Successfully benchmarked the method using generated test datasets derived from larger ones.

Conclusions:

• Public microarray databases can be effectively integrated into analysis pipelines for small sample-sized experiments.
• The developed approach offers a robust solution for identifying DEGs, particularly when sample sizes are limited.
• This work provides evidence supporting the utility of public data for advancing microarray data analysis.