Genomic basis of seed colour in quinoa inferred from variant patterns using extreme gradient boosting.

Quinoa is an agriculturally important crop species originally domesticated in the Andes of central South America. One of its most important phenotypic traits is seed colour. Seed colour variation is determined by contrasting abundance of betalains, a class of strong antioxidant and free radicals scavenging colour pigments only found in plants of the order Caryophyllales. However, the genetic basis for these pigments in seeds remains to be identified. Here we demonstrate the application of machine learning (extreme gradient boosting) to identify genetic variants predictive of seed colour. We show that extreme gradient boosting outperforms the classical genome-wide association approach. We provide re-sequencing and phenotypic data for 156 South American quinoa accessions and identify candidate genes potentially controlling betalain content in quinoa seeds. Genes identified include novel cytochrome P450 genes and known members of the betalain synthesis pathway, as well as genes annotated as being involved in seed development. Our work showcases the power of modern machine learning methods to extract biologically meaningful information from large sequencing data sets.

A straight-forward seed production technology system for foxtail millet (Setaria italica).

For autogamous crops, a precondition for using heterosis is to produce sufficient pure male-sterile female parents that can be used to produce hybrid seeds. To date, cytoplasmic male sterility (CMS) and environment-sensitive genic male sterility (EGMS) have been used commercially to exploit heterosis for autogamous species. However, neither CMS nor EGMS has been established for foxtail millet (Setaria italica). Here, we report on the establishment and application of a seed production technology (SPT) system for this crop. First, we established a DsRed-based SPT system, but found that it was unsuitable because it required the use of a fluorescent device for seed sorting. Instead, we constructed an SPT system with de novo betalain biosynthesis as the selection marker. This allowed us to distinguish transgenic seeds with the naked eye, thereby facilitating the identification of SPT maintainer line seeds. In this system, a seed sorter was not required to obtain sufficient seeds. The key point of the strategy is that the seed pool of the SPT maintainer line is propagated by artificial identification and harvesting of male-fertile individuals in the field, and the male-sterile line seed pool for hybrid production is produced and propagated by free pollination of male-sterile plants with the SPT maintainer line. In a field experiment, we obtained 423.96 kg male-sterile line seeds per acre, which is sufficient to plant 700.18 acres of farmland for hybrid seed production or male-sterile line reproduction. Our study therefore describes a powerful tool for hybrid seed production in foxtail millet, and demonstrates how the SPT system can be used for a small-grained crop with high reproduction efficiency.