Genomic prediction with NetGP based on gene network and multi-omics data in plants

Area of Science:

• Plant breeding
• Genomics
• Bioinformatics

Background:

• Genomic selection (GS) is a breeding strategy that uses whole-genome information for trait prediction.
• Traditional GS models have limited prediction accuracy due to challenges in capturing complex genotype-trait interactions.
• There is a need for advanced methods to improve prediction accuracy in plant breeding.

Purpose of the Study:

• To introduce a novel single nucleotide polymorphism (SNP) feature extraction technique, Pearson-Collinearity Selection (PCS).
• To develop a deep learning gene network prediction model (NetGP) for enhanced phenotypic prediction.
• To evaluate the performance of NetGP using genomic, transcriptomic, and multi-omics data.

Main Methods:

• Developed and applied the Pearson-Collinearity Selection (PCS) method for SNP feature extraction.
• Implemented a deep learning model, NetGP, utilizing transcriptomic (Trans) and genomic (SNP) datasets.
• Evaluated NetGP's performance on genomic, transcriptomic, and combined multi-omics (Trans + SNP) predictions.

Main Results:

• The PCS technique improved prediction accuracy across existing GS models.
• NetGP demonstrated superior performance in genomic, transcriptomic, and multi-omics predictions compared to other models.
• The NetGP multi-omics model outperformed independent genomic or transcriptomic prediction models and showed good generalizability across different plant datasets.

Conclusions:

• The PCS feature extraction and NetGP model offer a novel and effective tool for plant genomic selection.
• This approach significantly enhances prediction accuracy by better modeling genotype-trait relationships.
• The study opens new research avenues for advancing plant breeding through multi-omics data integration and deep learning.