Estimation of Complex-Trait Prediction Accuracy from the Different Holo-Omics Interaction Models.

Area of Science:

• Genetics and Bioinformatics
• Microbiome Research
• Quantitative Genetics

Background:

• Statistical models are crucial for breeding programs targeting complex traits.
• The holo-omics framework integrates host genome and microbiome data for trait prediction, but presents challenges.
• Combining genomic and microbiome data is essential for enhancing prediction accuracy.

Purpose of the Study:

• To evaluate prediction accuracy of holo-omics interaction models.
• To compare different methods for combining host genome and microbiome data.
• To investigate the holo-omics architecture of complex traits.

Main Methods:

• Developed and validated holo-omics interaction models using two public datasets.
• Compared holo-omics models against purely genomic and microbiome prediction models.
• Estimated prediction accuracy for eleven complex traits.

Main Results:

• Holo-omics interaction models achieved the highest prediction accuracy for ten out of eleven traits.
• The Hadamard product-based holo-omics interaction matrix was most accurate for nine traits.
• Direct holo-omics and microbiome models were superior for the remaining two traits.

Conclusions:

• Holo-omics models, particularly those using Hadamard product for interaction matrices, enhance prediction accuracy for complex traits.
• The study provides insights into the holo-omics architecture underlying complex traits.
• Integrating multi-omics data is a promising strategy for improving breeding programs.