Classification approaches for sorting maize (Zea mays subsp. mays) haploid using single-kernel near-infrared spectroscopy

Authors: GUSTIN, JEFFREY - University Of Florida; FREI, URSULA - Iowa State University; BAIER, JOHN - University Of Florida; Armstrong, Paul; LUBBERSTEDT, THOMAS - Iowa State University; SETTLES, MARK - University Of Florida

Submitted to: Plant Breeding
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/23/2021
Publication Date: 8/24/2020
Citation: Gustin, J.L., Frei, U.K., Baier, J., Armstrong, P.R., Lubberstedt, T., Settles, A.M. 2020. Classification approaches for sorting maize (Zea mays subsp. mays) haploid using single-kernel near-infrared spectroscopy. Plant Breeding. 139(6):1103-1112. https://doi.org/10.1111/pbr.12857.
DOI: https://doi.org/10.1111/pbr.12857

Interpretive Summary: Doubled haploids (DHs) seeds have become an important breeding tool for creating maize inbred lines. However, several bottlenecks in the DH production process limit wider development, application, and adoption of the technique. Haploid kernels are typically sorted manually from a much larger pool of hybrid siblings which introduces time constraints on DH production. Automated sorting based on the chemical composition of the kernel can be effective but have not achieved the necessary sorting speed to be cost-effective replacement over manual sorting. Single kernel near-infrared reflectance (skNIR) spectroscopy was evaluated as a platform to accurately identify haploid kernels. The skNIR platform is a high-throughput device that acquires a NIR spectrum and weight from each kernel to sort DH from hybrid kernels. With this system we were able to enrich the haploid selection pool to above 50% haploids which would make the final manual sort be performed on a substantially smaller lot of kernels.

Technical Abstract: Doubled haploids (DHs) have become an important breeding tool for creating maize inbred lines. However, several bottlenecks in the DH production process limit wider development, application, and adoption of the technique. Haploid kernels are typically sorted manually from a much larger pool of hybrid siblings in a haploid induction cross, which introduces several constraints on DH production. Automated sorting based on the chemical composition of the kernel can be effective, but proposed devices have not achieved the necessary sorting speed to be a cost-effective replacement to manual sorting. We evaluated the ability of a single kernel near-infrared reflectance spectroscopy (skNIR) platform to identify haploid kernels. The skNIR platform is a high-throughput device that acquires a NIR spectrum and weight from each kernel. We used skNIR data from 15 haploid induction crosses to construct general discrimination models based on all induction crosses and specific models that only considered kernels within an induction cross. Specific models outperformed the general model and were able to enrich a haploid selection pool to above 50% haploids. Potential applications for the device are discussed.