BLMA

BLMA performs bi-level meta-analysis to integrate genomic data from high-throughput studies, addressing study heterogeneity, bias, and outliers to increase robustness and sensitivity.

Key Features:

• Bi-Level Framework: Performs both intra-experiment and inter-experiment analyses using the additive method and the Central Limit Theorem to evaluate signals at two hierarchical levels.
• Robustness Against Bias and Outliers: Is less affected by outliers and bias compared with P-value-based meta-analysis techniques such as Fisher's, Stouffer's, minP, and maxP.
• Enhanced Sensitivity: Detects small changes in signal with greater sensitivity than several traditional methods.
• Comparative Power: The intra-experiment analysis demonstrates greater statistical power than equivalent tests conducted on aggregated large experiments, aiding pathway and differential expression analyses.
• Pathway Analysis Performance: In a comparative study of 1252 samples from 21 datasets related to acute myeloid leukemia, type II diabetes, and Alzheimer's disease, BLMA outperformed Fisher's, Stouffer's, the additive method, and the pathway meta-analysis package MetaPath in identifying relevant pathways.
• General Applicability: The framework can be applied across different types of statistical meta-analyses for diverse genomic datasets.

Scientific Applications:

• Differential Expression Analysis: Facilitates identification of differentially expressed genes across multiple independent experiments by leveraging bi-level statistical aggregation.
• Functional and Pathway Analysis: Improves detection and ranking of phenotype-relevant pathways in pathway-level meta-analyses across heterogeneous datasets.

Methodology:

BLMA applies a bi-level meta-analysis by performing intra-experiment statistical tests using the additive method and the Central Limit Theorem, then aggregates those results across independent experiments.