Modeling allele-specific expression at the gene and SNP levels simultaneously by a Bayesian logistic mixed regression...

Area of Science:

• Genomics
• Bioinformatics
• Statistical Genetics

Background:

• High-throughput sequencing generates vast data for allele-specific expression (ASE) analysis.
• Existing statistical methods for ASE are often simplistic and fail to capture gene-level and exon-level variations.
• There is a need for advanced statistical approaches to analyze complex gene expression patterns.

Purpose of the Study:

• To develop a novel statistical model for comprehensive ASE analysis.
• To simultaneously assess ASE for a gene as a whole and variations within a gene across exons.
• To improve the reliability and power of statistical inferences in ASE studies.

Main Methods:

• A generalized linear mixed model incorporating gene, SNP, and biological replicate variations.
• Bayesian model selection for testing ASE hypotheses at gene and SNP levels.
• Application to bovine tissue data for de novo ASE gene detection.
• Development of an R package (BLMRM) for public accessibility.

Main Results:

• The proposed method effectively detects allele-specific expression (ASE) in bovine genomes across four tissue types.
• Identified regulatory ASEs across gene exons and tissue types.
• Simulation studies demonstrated superior performance compared to existing methods.
• The BLMRM package provides a robust tool for ASE analysis.

Conclusions:

• The novel method offers improved control of false discovery rates and enhanced power for ASE detection.
• It effectively handles SNP and biological variations.
• The approach is computationally efficient, enabling whole-genome analysis.