Integrative Bayesian variable selection with gene-based informative priors for genome-wide association studies.

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Genome-wide association studies (GWAS) traditionally use univariate analysis for identifying phenotype-associated single nucleotide polymorphisms (SNPs).
• Current GWAS methods often fail to explain the full heritability of complex diseases, leading to the "missing heritability" problem.
• Gene interrelationships are often overlooked in standard GWAS analyses.

Purpose of the Study:

• To introduce and evaluate a novel integrative Bayesian Variable Selection (iBVS) strategy for GWAS.
• To address the "missing heritability" by incorporating gene network information into genetic analysis.
• To compare the performance of iBVS against existing methods like LASSO and Stepwise regression.

Main Methods:

• Implementation of an integrative Bayesian Variable Selection (iBVS) strategy.
• Development of a hierarchical model with an informative prior based on gene network interrelationships.
• Application of iBVS to simulated data and a real-world leprosy case-control study.
• Comparison of iBVS with Stepwise and LASSO methods using Area Under the Curve (AUC) for variable selection and outcome prediction.

Main Results:

• Simulation studies showed iBVS achieved the highest AUC for both variable selection and outcome prediction compared to Stepwise and LASSO.
• In a leprosy study, iBVS identified 94 SNPs, LASSO identified 100 SNPs, and Stepwise identified 3 SNPs.
• iBVS demonstrated comparable prediction performance to LASSO and superior performance to Stepwise regression.

Conclusions:

• The proposed iBVS strategy is a novel and effective method for genome-wide association studies.
• iBVS offers more interpretable posterior probabilities for variables compared to LASSO and other penalized regression methods.
• This approach provides a valuable tool for dissecting the genetic architecture of complex diseases.