Location: Stored Product Insect and Engineering Research
Project Number: 3020-43440-010-009-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Jun 15, 2024
End Date: Jun 14, 2026
Objective:
The primary objective of this work will be to evaluate the use of near infrared (NIR) spectroscopy to efficiently and accurately sort haploid from hybrids kernels based on oil content when compared with the traditional method of manually sorting kernels based on the R1-nj color marker difference between these types. Additional work will continue on the use of new material from collaborators which will have enhanced oil separation between the oil kernel types and thus potential for improved sorting accuracy. We will also explore the use of hyperspectral imaging and artificial intelligence to identify grain contaminated with mold, mycotoxins, or fungi in bulk grain samples.
Approach:
Haploid breeding can effectively accelerate maize breeding and is of great significance for improving breeding efficiency and saving costs. However, haploid kernels only occur in a small percentage, nominally 10%, of the kernels from the different lines of interest to breeders. Traditional haploid induction methods use a R1-nj color marker causing haploid and hybrid kernels be visually different and thus can be manually sorted. Manual sorting, however, can create bottlenecks in the breeding cycle. The approach is to use NIR to sort kernels based on the inherent oil content differences between haploid and hybrid kernels. ARS has developed the hardware and software for single seed NIR sorting and more recently applied adaptive sorting methods. Adaptive sorting varies hard thresholds, during sorting, to maintain certain percentages of kernels being routed into different classes. This utilizes the natural distribution of haploid and hybrid kernels in a breeding line. With adaptive sorting, sorting is still based on oil content, but sorting thresholds are varied during sorting to maintain a user-specified percentage distribution of kernels. Preliminary adaptive sorting into three streams has shown about 75% of the undesired hybrid seeds can be eliminated, a highly enriched haploid pool of 10% can be obtained, and a small mid-stream of mixed of types remain. Continued work by cooperator and collaborators to create greater oil differences, via a haploid inducer, should further improve what are already fairly acceptable separations of the kernel types.
