A bayesian mixed regression based prediction of quantitative traits from molecular marker and gene expression data

Area of Science:

• Genomics
• Bioinformatics
• Plant Pathology

Background:

• Predicting pathogen activity in soybean is crucial for crop management.
• Leveraging both molecular marker and gene expression data offers a comprehensive approach.

Purpose of the Study:

• To improve prediction accuracy of soybean pathogen-activity-phenotypes.
• To evaluate the effectiveness of combining different data types (molecular markers and gene expression).
• To explore variable selection strategies within Bayesian hierarchical models.

Main Methods:

• Bayesian hierarchical regression modeling was employed for phenotype prediction.
• Molecular marker and gene expression data were used as additive predictors.
• Various predictor selection strategies and cross-validation were utilized to assess prediction accuracy.

Main Results:

• Simultaneous use of functional genomic and genetic data significantly improved out-of-data prediction accuracy.
• Complex models achieved adequate goodness-of-fit due to a large number of potential predictors and sufficient sample size.
• Gene-set and chromosomal enrichment analyses provided further biological insights.

Conclusions:

• Integrating diverse genomic data types enhances predictive performance in soybean.
• Careful variable selection is essential for mitigating overfitting in complex biological models.
• Bayesian hierarchical modeling with indicator-based covariate selection offers a robust framework for genetic prediction.