Evaluating reproducibility of differential expression discoveries in microarray studies by considering correlated...

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Microarray studies for complex diseases often yield inconsistent lists of differentially expressed genes (DEGs).
• Traditional consistency metrics like percentage of overlapping genes (POG) fail to capture coordinated molecular changes.
• This irreproducibility problem extends to other high-throughput omics fields like proteomics and metabolism.

Purpose of the Study:

• To develop novel metrics for evaluating the consistency of DEG lists from different studies.
• To address the limitations of existing methods in assessing reproducibility for complex diseases.
• To improve the interpretation of high-throughput omics data.

Main Methods:

• Proposed new metrics: percentage of overlapping genes-related (POGR) and normalized POGR (nPOGR).
• These metrics consider correlated molecular changes, not just direct gene overlaps.
• Applied metrics to microarray datasets from three complex diseases.

Main Results:

• DEG lists from different studies showed low POG scores but high POGR and nPOGR scores.
• This indicates significant underlying correlation and reproducibility despite apparent inconsistencies.
• The proposed metrics effectively highlight reproducibility missed by POG.

Conclusions:

• The POGR and nPOGR metrics offer a more accurate assessment of DEG list reproducibility in complex diseases.
• These metrics can reduce uncertainty in interpreting microarray and other omics study results.
• The approach is applicable across various high-throughput post-genomic research areas.