METRADISC

METRADISC performs ranked meta-analysis of discovery-oriented biological datasets to integrate multidimensional biological signals, such as gene expression, and assess between-study heterogeneity for each biological variable.

Key Features:

• Generalized meta-analysis methodology: METRADISC employs a generalized meta-analysis framework to integrate information across multiple datasets.
• Non-parametric Monte Carlo permutation testing: It uses non-parametric Monte Carlo permutation testing to assess significance of ranked variables.
• Rank-based analysis: Variables are ranked by statistical significance within each study and metrics are derived from those ranks.
• Average-rank and heterogeneity metrics: The tool evaluates average rank and between-study heterogeneity of ranks for each variable.
• Permutation-based null distribution generation: It generates null distributions for average-rank and heterogeneity metrics by randomly permuting ranks within each study, accounting for ties and varying numbers of tested variables.

Scientific Applications:

• Integration of large-scale discovery datasets: Combine complex datasets from massive testing efforts to synthesize findings across studies.
• Gene expression meta-analysis in prostate cancer: Application demonstrated on seven prostate cancer gene expression studies comparing cases with normal controls.
• Identification of consistently significant biological variables: Identify variables that consistently show significance across multiple studies.
• Assessment of result consistency and diversity: Examine consistency and diversity of results across different studies to detect genuine biological heterogeneity.

Methodology:

Rank variables within each study, compute average-rank and heterogeneity metrics across studies, and perform non-parametric Monte Carlo permutation testing by randomly permuting ranks within each study (accounting for ties and variations in tested variables) to generate null distributions for significance assessment.