Using drone-retrieved multispectral data for phenomic selection in potato breeding

Area of Science:

• Plant breeding
• Agricultural science
• Genomics and phenomics

Background:

• Potato breeding programs face challenges with large numbers of entries and limited tubers in early stages.
• Predictive breeding strategies like genomic selection (GS) and phenomic selection (PS) offer potential solutions.

Purpose of the Study:

• To evaluate drone-derived multispectral reflectance for phenomic prediction in diverse potato material across various traits.
• To compare the predictive performance of phenomic versus genomic selection.
• To assess the utility of mixed relationship matrices combining SNP and spectral data.

Main Methods:

• Testing various phenomic prediction scenarios using drone-based multispectral data on tetraploid potato panels.
• Comparing predictive abilities of phenomic and genomic selection models.
• Utilizing mixed relationship matrices integrating SNP array and multispectral reflectance data.

Main Results:

• Phenomic prediction accuracy varied widely (–0.15 to 0.88), influenced by environment, trait, and scenario.
• High predictive abilities were observed for yield, maturity, foliage development, and emergence using phenomic data.
• Mixed models integrating both genomic and phenomic data improved prediction for 20 out of 22 traits, indicating complementary information.

Conclusions:

• Phenomic selection is valuable for early-stage potato breeding (seedling/single hill) where high-throughput genotyping is cost-prohibitive.
• Drone-derived spectral data can effectively support phenomic prediction for specific potato traits.
• Integrating genomic and phenomic data offers a powerful approach to enhance selection gain in potato breeding programs.