MT-HESS: an efficient Bayesian approach for simultaneous association detection in OMICS datasets, with application to...

Area of Science:

• Integrative genomics
• Statistical genetics
• Bioinformatics

Background:

• Analyzing associations between genetic variants (predictors) and gene expression (responses) is crucial in genomics.
• Identifying 'hotspots'—genetic variants regulating many genes—is of particular interest.
• Determining if expression Quantitative Trait Loci (eQTLs) are shared across tissues/conditions or are condition-specific is a key research question.

Purpose of the Study:

• To introduce MT-HESS, a Bayesian hierarchical model for joint analysis of multiple predictors and responses across diverse conditions.
• To improve the detection of genetic hotspots by leveraging shared information across genes.
• To enhance statistical power for identifying eQTLs compared to traditional methods.

Main Methods:

• Implemented a Bayesian hierarchical model (MT-HESS) for multivariate analysis of SNP-gene expression associations.
• Employed a Bayesian sparse regression algorithm for comprehensive model search across linear combinations of SNPs.
• Incorporated a model for correlations between condition/tissue-specific responses and a hierarchical structure for gene-wise information sharing.

Main Results:

• Demonstrated increased statistical power through extensive simulations.
• Identified novel genetic hotspots in case studies that were missed by standard approaches.
• Showcased improved prediction power by jointly analyzing multiple tissues.

Conclusions:

• MT-HESS provides a powerful framework for integrative genomics analysis, particularly for identifying shared and condition-specific eQTLs.
• The model effectively leverages multi-tissue data to detect genetic hotspots and enhance prediction accuracy.
• This approach advances the understanding of genetic regulation of gene expression across different biological contexts.