A Bayesian random regression method using mixture priors for genome-enabled analysis of time-series high-throughput...

Area of Science:

• Plant genetics and breeding
• Quantitative genetics
• Bioinformatics and computational biology

Background:

• Image-based phenotyping platforms generate large-scale, time-series crop data.
• Efficiently utilizing longitudinal phenotypic data is crucial for understanding genetic architecture and improving plant breeding.
• Existing methods may not fully capture the dynamic nature of crop traits over time.

Purpose of the Study:

• To develop a Bayesian random regression model (RR-BayesC) for analyzing longitudinal crop traits.
• To incorporate mixture priors for marker effects to improve biological assumptions in genomic analyses.
• To evaluate the performance of RR-BayesC for genomic prediction and genome-wide association studies (GWASs) using simulated and real rice data.

Main Methods:

• Development of a Bayesian random regression model (RR-BayesC) integrating mixture priors for marker effects (Bayes Cπ).
• Application of the model to analyze simulated and real rice (Oryza sativa L.) data.
• Evaluation of genomic prediction accuracy across different scenarios and GWAS performance for identifying time-variant and time-invariant quantitative trait loci (QTLs).

Main Results:

• RR-BayesC demonstrated significantly higher prediction accuracy than single-trait analysis in simulated data, especially when heritability was low.
• In real data, RR-BayesC achieved high prediction accuracy for forecasting later-stage phenotypes, even for lines with limited early observations.
• Simulated GWASs showed RR-BayesC could effectively distinguish QTLs invariant to time from those interacting with time.

Conclusions:

• The developed Bayesian random regression model (RR-BayesC) effectively analyzes longitudinal crop data.
• RR-BayesC enhances the accuracy of genomic prediction and the power of GWASs for dynamic crop traits.
• The associated software tool JWAS facilitates the application of this advanced random regression analysis in plant breeding.