A Bayesian framework for inference of the genotype-phenotype map for segregating populations

Area of Science:

• Genetics
• Systems Biology
• Computational Biology

Background:

• Complex genetic interactions underpin many diseases, necessitating efficient methods for their study.
• In vivo hypothesis testing in mammals is resource-intensive and limited in scope.
• Computational approaches are vital for prioritizing genetic targets for disease intervention.

Purpose of the Study:

• To develop a Bayesian statistical method for inferring causal networks from genotype and phenotype data.
• To utilize expression quantitative trait loci (eQTL) data from genetically randomized populations.
• To enable prediction of system-wide responses to genetic perturbations.

Main Methods:

• A Bayesian approach employing hierarchical regression models to define causal relationships.
• Incorporation of prior biological knowledge through network structure priors.
• Utilizing an efficient Monte Carlo method for network structure searching and sampling.
• Application to kidney gene expression data from an MRL/MpJ × SM/J intercross population.

Main Results:

• Inference of a network of causal relationships between genotypes and phenotypes.
• Generation of an ensemble of networks providing confidence in topology.
• Prediction of system-wide responses to perturbations.
• Identification of a previously uncharacterized feedback loop in the local renin-angiotensin system.

Conclusions:

• The proposed Bayesian method effectively infers causal genetic networks from eQTL data.
• This approach facilitates the prioritization of targets for understanding disease mechanisms.
• The identified feedback loop offers a new avenue for research into kidney function and disease.