bacon

bacon implements a Bayesian method to control bias and inflation in epigenome-wide (EWAS) and transcriptome-wide (TWAS) association studies by estimating an empirical null distribution from study z-scores.

Key Features:

• Bias and inflation control: Controls bias and inflation in test statistics specific to EWAS and TWAS.
• Empirical null estimation: Estimates an empirical null distribution from z-scores derived from study data.
• Gibbs Sampling mixture model: Fits a three-component normal mixture model to z-scores using Gibbs Sampling.
• Statistical performance: Demonstrates increased statistical power while controlling the false positive rate in evaluations.
• Validation: Validated through simulations and applications to real-world datasets.
• Meta-analysis applicability: Applicable to large-scale meta-analyses and datasets examining complex traits such as age and smoking.

Scientific Applications:

• Bias correction in EWAS/TWAS: Reduces spurious findings by adjusting for bias and inflation in epigenome- and transcriptome-wide association studies.
• Improved association detection: Enhances identification of true associations in large-scale EWAS and TWAS datasets.
• Meta-analysis of complex traits: Supports meta-analyses of traits such as age and smoking where inflation and bias are common.

Methodology:

Estimates an empirical null by fitting a three-component normal mixture model to study z-scores using Gibbs Sampling.