Doubled haploid (DH) technology significantly shortens the breeding cycle for developing homozygous inbred lines in maize (Zea mays). Manual sorting of haploids from a larger bulk of hybrid kernels in an induction cross is a major bottleneck in DH development. Automated systems based on near-infrared (NIR) reflectance spectroscopy can be valuable tools for rapid haploid sorting, provided that sorting accuracy is sufficient for incorporation into the DH process. In this study, we evaluated the accuracy of a custom-built single-kernel NIR (skNIR) sorter for classifying haploid kernels from 12 high-oil haploid induction populations generated from two sweet corn and two field corn donors and four high-oil haploid inducers (HOHIs). We evaluated several general classification models that can be applied without population-specific recalibration or prior genotyping, including models that classified haploids based solely on predicted oil content, as well as multivariate methods that used all wavelengths of the NIR spectra. The highest classification accuracy was obtained using a general multivariate support vector machine (SVM) model. When combined with the two best-performing HOHIs, the general SVM model accurately sorted induction populations from two of the three donor backgrounds crossed with these inducers. Two oil-based methods showed less accurate classification than the multivariate SVM model, due to overlapping oil content distributions across the two kernel classes. Overall, this study demonstrates effective skNIR sorting of haploid kernels from diverse induction populations using a single general model. The practical deployment of this instrument in maize breeding programs is discussed.

In this study, we evaluated a single-kernel near-infrared (NIR) spectroscopy sorter as an automated, pigmentation-independent alternative for haploid classification across 12 induction populations comprising two field corn and two sweet corn donors crossed with four HOHIs. Two complementary sorting approaches were assessed: oil-based methods that exploit the relative difference in oil content between hybrid and haploid kernels (Δ oil) produced with HOHIs and multivariate methods that classify kernels using all wavelengths of the skNIR spectra (940-1640 nm) rather than a single predicted trait.

NIR spectra were collected from kernels of 12 haploid induction populations using the skNIR sorter. Individual kernels were selected from the 24-well microtiter plate and manually dropped into an illuminated light tube. Fiber optic cables at the top and bottom of the light tube transmitted light to a CDI 256L-1.7T1 spectrophotometer, which features a 256-element InGaAs photodiode array spanning the spectral range from 906 to 1679 nm with a 1 nm step size. NIR reflectance values were trimmed to 940 to 1640 nm, and absorbance values were calculated as log(1/Reflectance). Oil content predictions for each kernel were calculated using partial least squares regression (PLSR) coefficients from a PLS model that was developed on the skNIR sorter using a well-established chemometric approach. Reference kernel oil content was determined for 9 of the 12 induction populations using a PCT-20/20B Nuclear Magnetic Resonance analyzer. 

For each induction cross, kernels were visually scored as they were shelled from the ears into normal (typical grain-fill, embryo and scutellum development) or defective. Normal kernels were visually scored as haploid or hybrid using the R1-nj color marker. After skNIR data collection, 192 haploid and hybrid kernels per population were randomly selected for genotyping using gDNA markers. Kernels were germinated directly in their resident 24-well plate using hydrogel beads. Eighty-two of the selected kernels did not germinate and were not used in the study. Coleoptile and radicle tissue from each germinated kernel were used for DNA extraction. Ploidy was confirmed for each germinated kernel based on high-resolution melting curve analysis using primer pairs that discriminate between the wild type (donor) and mutated (inducer) alleles for the haploid induction gene MATRILINEAL. Of the 2,403 total kernels in the 12 induction populations, 393 were classified as haploid and 1,829 as hybrid by gDNA scoring, and 181 were classified as defective based on visual scoring.