Meta-analysis for ranked discovery datasets: theoretical framework and empirical demonstration for microarrays

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Combining large-scale biological datasets, like gene expression profiles, poses significant challenges.
• Existing methodologies struggle to efficiently integrate diverse datasets while assessing inter-study heterogeneity.
• There is a need for robust tools to analyze multidimensional biological signals across multiple studies.

Purpose of the Study:

• To introduce METa-analysis of RAnked DISCovery datasets (METRADISC), a generalized meta-analysis method.
• To enable efficient combination of information from discovery-oriented datasets.
• To provide a framework for testing between-study heterogeneity for biological variables.

Main Methods:

• METRADISC employs non-parametric Monte Carlo permutation testing.
• Biological variables are ranked by statistical significance within each study.
• The method assesses the average rank and between-study heterogeneity of ranks using permuted data.

Main Results:

• METRADISC successfully combines information from multiple discovery datasets.
• The method effectively tests for heterogeneity in biological variables across studies.
• Empirical validation using prostate cancer gene expression data demonstrated its utility.

Conclusions:

• METRADISC offers a new computational tool for meta-analysis of large-scale biological data.
• The method facilitates the examination of result diversity across combined studies.
• This approach enhances the analysis of complex datasets from discovery-oriented research.