Integration of multi-omics data for prediction of phenotypic traits using random forest

Area of Science:

• Plant genetics and breeding
• Metabolomics and proteomics
• Bioinformatics and systems biology

Background:

• Understanding genetic and metabolic pathways is crucial for potato tuber quality.
• Multi-omics data (gene expression, metabolites, proteins) offer a comprehensive view.
• Integrating diverse omics datasets can improve phenotypic trait prediction.

Purpose of the Study:

• To develop and present an approach for integrating multi-omics data in potato.
• To predict phenotypic traits related to tuber quality.
• To identify networks of omics variables associated with specific traits.

Main Methods:

• Analysis of gene expression, GC-MS and LC-MS metabolite, and proteomics data.
• Utilized Random Forest regression for data integration and trait prediction.
• Employed Expression QTL (eQTL) analysis for identifying genomic associations.

Main Results:

• Random Forest successfully predicted tuber flesh color, DSC onset, tuber shape, and enzymatic discoloration.
• Gene expression data was key for predicting tuber shape.
• Integration of multi-omics data explained up to 75% of variation for tuber flesh color.
• Identified significant genes, metabolites, and proteins associated with DSC onset and enzymatic discoloration.

Conclusions:

• A strategy for selecting and integrating multi-omics data using Random Forest was established.
• Network analysis revealed associations between gene expression, metabolites, and proteins for specific traits.
• The approach allows for accurate prediction of potato tuber quality traits.