Genomic prediction powered by multi-omics data.

Area of Science:

• Plant breeding and genetics
• Bioinformatics and computational biology

Background:

• Genomic selection (GS) predicts complex plant traits but is limited by genomic data alone.
• Integrating transcriptomics and metabolomics offers a more comprehensive view of molecular mechanisms for enhanced prediction.

Purpose of the Study:

• To assess the effectiveness of 24 multi-omics integration strategies for improving genomic prediction accuracy.
• To compare early data fusion (concatenation) with model-based integration techniques.

Main Methods:

• Utilized three real-world datasets from maize and rice under single-environment conditions.
• Evaluated 24 strategies combining genomics, transcriptomics, and metabolomics using concatenation and model-based fusion.
• Assessed predictive performance for complex traits, considering population size and omics dimensionality.

Main Results:

• Model-based multi-omics integration consistently improved predictive accuracy over genomic-only models, especially for complex traits.
• Simple concatenation strategies did not consistently enhance prediction and sometimes underperformed.
• The choice of integration strategy significantly impacts the success of multi-omics for genomic prediction.

Conclusions:

• Sophisticated model-based integration is crucial for maximizing the benefits of multi-omics data in plant breeding.
• Appropriate strategy selection is key to leveraging multi-omics for accelerated genetic gain.
• This study provides practical insights for designing effective omics-informed selection strategies.