CNPBayes

CNPBayes applies Bayesian hierarchical Gaussian mixture models and hidden Markov models to detect and estimate copy number polymorphisms (CNPs) from high-throughput SNP array data and test their associations with phenotypes such as estimated glomerular filtration rate (eGFR), while accommodating batch effects.

Key Features:

• Bayesian Hierarchical Gaussian Mixture Models: Estimates copy numbers at genomic regions using Bayesian Gaussian mixture models to provide probabilistic copy-number assignments and accommodate uncertainty.
• Hidden Markov Models (HMMs): Identifies CNP regions from high-throughput SNP array data using HMMs.
• Batch-effect accommodation: Models and mitigates batch effects inherent to high-throughput SNP array analyses via the hierarchical Bayesian framework.
• Population-Specific Analysis: Performs separate analyses for ancestry groups such as African American (AA) and European Ancestry (EA) to capture population-specific CNPs.
• Adjustment for Population Structure: Incorporates multivariate models adjusted for SNP-derived covariates of population structure to reduce ancestry-related confounding.
• Multiple-testing Correction: Assesses statistical significance with adjustments for multiple comparisons, for example Bonferroni correction.
• Genomic-region Association Detection: Detects associations at specific genomic loci, as illustrated by signals near chromosome 5 in EA cohorts.

Scientific Applications:

• Genetic Association Studies: Tests associations between CNPs and quantitative phenotypes such as eGFR to identify genetic contributors to kidney function.
• Exploration of Genetic Variability: Extends GWAS beyond SNPs by enabling evaluation of copy number polymorphisms as contributors to complex trait variation.

Methodology:

Input high-throughput SNP array data; identify CNP regions with hidden Markov models; estimate copy numbers using Bayesian Gaussian mixture models within a hierarchical framework that accounts for batch effects; adjust associations using multivariate models with SNP-derived population-structure covariates; assess significance with multiple-comparison correction (e.g., Bonferroni).