Integration of proxy intermediate omics traits into a nonlinear two-step model for accurate phenotypic prediction

Area of Science:

• Plant biology
• Genomics
• Microbiome research
• Metabolomics

Background:

• Intermediate omics traits are crucial for understanding genetic effects on phenotypes.
• Rhizosphere microbiota significantly impact plant health, but host genetic interactions are complex.
• Existing two-step models for omics integration lack nonlinear relationship modeling.

Purpose of the Study:

• To develop a two-step phenotype prediction framework integrating genomic, microbiome, and metabolome data.
• To explicitly capture nonlinearities between different omics layers.
• To improve the accuracy of plant phenotype prediction and reduce reliance on costly omics data.

Main Methods:

• A two-step prediction framework was proposed: 1. Predict meta-metabolome traits from genetic and microbial data. 2. Use generated 'proxy' omics traits to enhance phenotype prediction.
• Linear mixed models (Best Linear Unbiased Prediction, BLUP) and nonlinear models (Random Forest, RF) were compared at each step.
• Simulations and a multi-omics soybean dataset were used for validation.

Main Results:

• The nonlinear modeling approach effectively captured intricate omics interactions.
• The proposed framework achieved improved phenotype prediction accuracy.
• Phenotype prediction was possible without original meta-metabolome training data, reducing measurement costs.

Conclusions:

• The novel framework successfully integrates intermediate omics traits into genomic prediction.
• Capturing omics-omics nonlinearities enhances prediction accuracy and provides deeper insights into plant-microbiome interactions.
• This approach offers a cost-effective solution for complex biological predictions.