Phenomic prediction of maize hybrids

Area of Science:

• Plant Science
• Agricultural Technology
• Biotechnology

Background:

• Large-scale phenomic experiments generate vast image data for plant development.
• Automated image processing converts raw data into genotype-feature matrices over time.

Purpose of the Study:

• To explore predicting mature plant biomass from early developmental stage image-derived features.
• To assess the utility of parental inbred lines' phenomic features for predicting hybrid biomass.
• To examine mid-parent heterosis in phenomic features.

Main Methods:

• Phenomic experiments conducted on 195 inbred and 382 hybrid maize varieties.
• Data collected from 16 to 48 days after sowing (DAS) using 129 image-derived features.
• Sparse regression methods applied to predict biomass from early-stage features.

Main Results:

• Approximately 20 early-stage features explained 73% of hybrid fresh weight variance and over 90% of dry weight variance.
• Using parental inbred line features predicted hybrid biomass with 42-45% variance explained.
• Phenomic prediction showed robustness across the experimental timescale.

Conclusions:

• Phenomic prediction is a viable alternative to genomic and metabolic prediction for hybrid performance.
• Early-stage phenomic data, especially from parental lines, is highly valuable for predicting hybrid biomass.
• Phenomic prediction offers a powerful tool for accelerating crop improvement.