BLMRM

BLMRM performs genome-wide detection and analysis of allele-specific expression (ASE) using a Bayesian logistic mixed regression framework to assess ASE at gene, exon, and single nucleotide polymorphism (SNP) levels.

Key Features:

• Model framework: Implements a Bayesian Logistic Mixed Regression Model within a generalized linear mixed model framework.
• Variation modeling: Integrates variation attributable to genes, SNPs, and biological replicates.
• Multi-level testing: Simultaneously tests ASE across entire genes, individual exons, and individual SNPs.
• Bayesian priors: Assigns priors on each model effect to facilitate information sharing across the genome.
• Bayesian model selection: Employs Bayesian model selection to test hypotheses of ASE per gene and for SNP-level variation within genes.
• Data compatibility: Designed for data derived from high-throughput sequencing experiments.
• Validation and performance: Demonstrates in simulation studies improved false discovery rate control and increased detection power when SNP variation and biological variability are present.
• Computational requirements: Maintains low computational requirements compatible with whole-genome analysis.

Scientific Applications:

• Genome-wide ASE discovery: Detection and analysis of allele-specific expression across the genome at gene, exon, and SNP resolution.
• Regulatory ASE mapping: Identification of regulatory ASEs across exons and tissues, as demonstrated in bovine tissue studies.
• Method benchmarking: Evaluation of ASE detection performance via simulation studies that mimic real sequencing datasets.

Methodology:

Uses a generalized linear mixed model implemented as a Bayesian Logistic Mixed Regression Model that models gene, SNP, and biological replicate effects, assigns Bayesian priors to each effect, and applies Bayesian model selection to test ASE per gene and SNP; validated using simulation studies that mimic real datasets.