Predicting the genetic component of gene expression using gene regulatory networks

Area of Science:

• Genomics
• Systems Biology
• Bioinformatics

Background:

• Gene expression prediction is crucial for transcriptome-wide association studies (TWAS).
• Traditional TWAS models primarily use local genetic variants.
• The omnigenic model suggests complex traits are influenced by all genetic variants.

Purpose of the Study:

• To develop a novel gene expression prediction approach.
• To incorporate distal genetic variants through gene regulatory networks (GRNs).
• To improve prediction accuracy for TWAS.

Main Methods:

• Reconstructed causal and coexpression Bayesian networks.
• Employed a two-step prediction process: local variants first, then distal variants via network.
• Utilized regularized regression for parameter estimation.

Main Results:

• The GRN-based approach outperformed traditional methods on simulated and real data (yeast, human).
• Successfully integrated both local and distal genetic influences for enhanced prediction.
• Demonstrated the utility of GRNs in modeling complex gene expression patterns.

Conclusions:

• Gene regulatory networks offer a powerful framework for gene expression prediction in TWAS.
• Incorporating distal genetic variants through GRNs enhances prediction accuracy.
• This method provides valuable insights for understanding complex trait genetics.