NBIA

NBIA integrates effect-size-based meta-analysis with gene-regulatory-network topology-aware network and pathway analysis to identify consistent genes, networks, and pathways across multiple expression studies.

Key Features:

• Effect-size focus: Emphasizes effect sizes across independent studies rather than only statistical significance to capture actual expression changes.
• Network topology integration: Incorporates the topological order and structural information of gene regulatory networks into analysis.
• Classical and modern meta-analytic techniques: Leverages both classical and modern meta-analysis methods in combination.
• Network-based pathway transformation: Transforms integrative analysis problems into standard pathway analysis problems using network-based approaches.
• Comparative benchmarking: Evaluated against nine other meta-analysis approaches including Impact Analysis, Gene Set Analysis (GSA), Gene Set Enrichment Analysis (GSEA) combined with Fisher's and the additive method, and three MetaPath approaches.
• Dataset-scale evaluation: Applied to 1,737 samples from 27 expression datasets spanning Alzheimer's disease, acute myeloid leukemia (AML), and influenza.
• Independent validation: Identified an AML signature validated on an independent cohort of 167 AML patients with a Cox p-value of 2 × 10^-6 distinguishing two patient groups with different survival.

Scientific Applications:

• Cross-study gene and network discovery: Identification of consistent gene- and network-level signals across multiple expression datasets.
• Pathway-level interpretation: Conversion of integrative results into pathway analysis for mechanistic interpretation of biological processes.
• Disease signature discovery: Detection of disease-relevant signatures in Alzheimer's disease, acute myeloid leukemia (AML), and influenza datasets.
• Clinical stratification and survival analysis: Derivation and validation of AML patient subgroups with significantly different survival profiles (Cox p-value 2 × 10^-6).
• Method benchmarking: Comparative assessment of meta-analysis approaches for robust identification of biologically relevant pathways and networks.

Methodology:

Integrate expression data across studies, compute effect-size-based meta-analyses using classical and modern methods, incorporate gene regulatory network topology into network-based analyses and pathway transformation, and validate signatures with survival analysis (Cox test); comparisons include Impact Analysis, GSA, GSEA with Fisher's and additive methods, and MetaPath approaches.