Jianbo He ¹ Lianshun Fu ² Xiaoshuai Hao ¹ Yicun Wu ¹ Mengfan Wang ¹ Qi Zhang ¹ Weidan Feng ¹ Mengmeng Fu ¹ Yanping Wang ³ Haixiang Ren ³ Weiguang Du ³ Wubin Wang ¹ Junyi Gai ^1*

• ¹ Nanjing Agricultural University, Nanjing, China
• ² Tieling Academy of Agricultural Sciences, Tieling, China
• ³ Mudanjiang Research and Development Center for Soybean/Mudanjiang Experiment Station of the National Center for Soybean Improvement, Mudanjiang Branch of Heilongjiang Academy of Agricultural Sciences, Mudanjiang, China

Northeast China (NEC) is the major production area for soybeans in China, whereas its soybean germplasm has played key roles in world soybean production, especially in Americas. For plant breeding, genomic selection involves two stages, cross design and progeny selection, with the former determines the latter's potential. In NEC, one of the major breeding purposes is for 100-seed weight (100SW) and seed oil content (SOC).A diverse sample with 361 NEC soybean germplasm accessions was evaluated for their 100SW and SOC in Tieling, Liaoning, China. Both traits exhibited significant phenotypic, genotypic, and G×E variation, with trait heritability of 82.38% and 86.26 %, respectively. A restricted two-stage multi-locus genome-wide association study (RTM-GWAS) with 15,501 SNPLDB (SNP linkage disequilibrium block) markers identified 80 and 92 QTLs with 230 and 299 alleles for 100SW and SOC, respectively.Corresponding to some increase of the two traits, almost all the alleles in the early maturity groups (MG 0+00+000) were inherited from the late MGs (MG I+II+III), indicating genetic recombination was the major motivator in addition to a few allele emergence and some allele exclusion fluctuations among early MGs. Using 95 th percentile as indicator, the prediction of recombination potentials showed that 30.43 g 100SW and 27.73% SOC might be achieved, respectively. Three strategies of simultaneous genomic improvement of both traits in designing optimal crosses, including 100SW-first, SOC-first, and 100SW-SOC-balance were proved efficient.Thus the optimal cross design could be extended to multiple traits based on a relatively thorough identification of the QTLs-alleles using RTM-GWAS.