**4. Discussion**

### *4.1. Phenotypic Variation of Regeneration Rate in Parent Lines and F1 Population*

The radish is a highly self-incompatible plant, and it is difficult to produce homozygous lines due to this characteristic. In previous studies, the parental lines with fixed traits through generation advancement were used for QTL analysis [34,35]. In cross-fertilization plants, bud pollination should be done by opening immature pollen and attaching pollen to mature flowers. Accordingly, it takes a lot of time and effort to develop lines with any traits fixed. In this study, in order to resolve this problem, a haploid breeding method has been attempted to obtain homozygous plants in a short period of time by doubling chromosomes [14,15]. Since the QTL analysis is based on allelic differences between parent lines, it is important to select parents that exhibit a broad range of phenotypic variations for the trait. Based on the comprehensive judgment, GX50 and GX71 were selected for crossing to generate an F1 mapping population, showing differences in terms of regeneration rate. The experiments showed that over 50% of the population did not regenerate, and only a few have high regeneration rates. These regeneration rates are consistent with previous findings [23,36].

### *4.2. GBS Analysis Using an NGS Platform and Genetic Mapping Using SNP Markers*

This study developed markers showing polymorphism between two parental lines, and genotyping for all F1 individuals was completed for a number of polymorphic markers. An next-generation sequencing (NGS) analysis is a sequencing method that can generate large amounts of genomic information at a low cost and is applied for genomic research of various plants. Unlike the costly and time-consuming re-sequencing, GBS analysis is a technique that uses a reduced representation library (RRL) to partially sequence a genome of interest using restriction enzymes [37]. Since GBS does not sequence the entire genome, it can reduce the complexity of the chromosome to allow a complex genome analysis, and large numbers of SNPs can be found for a variety of plant genetic studies, including genetic mapping and population analysis [23,38]. Creating a linkage map plays a

very important role in detecting the location of loci in the chromosome, such as QTL analysis. SNPs are the most frequently occurring genomic variations and can be used as markers for genome analysis, mapping, marker-assisted breeding, etc. The GBS method was adopted to obtain a large amount of SNPs in this study. In the preparation of the GBS library, a double-digestion method using *Nsi*l-HF and *MseI* was applied to reduce the genomic representation so that the sequence coverage for each allele increases. The library was generated with an average size of 344 bp. Considering that the size of the library suitable for NGS analysis is 170 ~ 350 bp, the generated library is considered suitable for sequencing [37]. Through the high filtering, 4462 high quality SNP markers were extracted and filtered again to generate 439.2 cM and 420.8 cM linkage maps using 178 markers, respectively. The average distance of the markers is 5.9 and 5.2 cM, respectively. When considering the length of the entire map, the interval between the markers is decent, but the length per linkage group is short [39,40]. The short linkage groups could be caused by two reasons: (1) the number of F1 individuals is small, and (2) we excluded redundant SNPs (could be reside in the same LD blocks). The genetic maps generated by the GBS have common lengthy linkage groups because some redundant markers overestimate the marker intervals. In this study, we tried to avoid those overestimated intervals by removing redundant SNPs. However, the number of markers mapped may be su fficient for a downstream analysis since we excluded duplicated markers in the same LD blocks. In addition, the SNP markers were mapped based on the locations in the pseudomolecule-level reference genome, and the physical position represented by the marker is clear, which may be useful for QTL mapping and maker-assisted selection (MAS).
