Identification of quantitative trait nucleotides and candidate genes for soybean seed weight by multiple models of genome-wide association study

BACKGROUND: Seed weight is a complex yield-related trait with a lot of quantitative trait loci (QTL) reported through linkage mapping studies. Integration of QTL from linkage mapping into breeding program is challenging due to numerous limitations, therefore, Genome-wide association study (GWAS) provides more precise location of QTL due to higher resolution and diverse genetic diversity in un-related individuals. RESULTS: The present study utilized 573 breeding lines population with 61,166 single nucleotide polymorphisms (SNPs) to identify quantitative trait nucleotides (QTNs) and candidate genes for seed weight in Chinese summer-sowing soybean. GWAS was conducted with two single-locus models (SLMs) and six multi-locus models (MLMs). Thirty-nine SNPs were detected by the two SLMs while 209 SNPs were detected by the six MLMs. In all, two hundred and thirty-one QTNs were found to be associated with seed weight in YHSBLP with various effects. Out of these, seventy SNPs were concurrently detected by both SLMs and MLMs on 8 chromosomes. Ninety-four QTNs co-localized with previously reported QTL/QTN by linkage/association mapping studies. A total of 36 candidate genes were predicted. Out of these candidate genes, four hub genes (Glyma06g44510, Glyma08g06420, Glyma12g33280 and Glyma19g28070) were identified by the integration of co-expression network. Among them, three were relatively expressed higher in the high HSW genotypes at R5 stage compared with low HSW genotypes except Glyma12g33280. Our results show that using more models especially MLMs are effective to find important QTNs, and the identified HSW QTNs/genes could be utilized in molecular breeding work for soybean seed weight and yield. CONCLUSION: Application of two single-locus plus six multi-locus models of GWAS identified 231 QTNs. Four hub genes (Glyma06g44510, Glyma08g06420, Glyma12g33280 & Glyma19g28070) detected via integration of co-expression network among the predicted candidate genes.