【Objective】To thoroughly dissect the QTL system conferring 100-seed weight in a recombinant inbred lines population, the restricted two-stage multi-locus genome-wide association analysis (RTM-GWAS) method was compared with other mapping methods for method optimization, which will provides basis for further exploration of candidate gene system and molecular marker-assisted design breeding. 【Method】A recombinant inbred line population consisting of 427 lines derived from a cross between Kefeng-1 and NN1138-2 was tested for its 100-seed weight under three environments. A total of 3 683 SNPLDBs (SNP linkage disequilibrium blocks) composed of 39 353 SNPs were applied to QTL mapping using three different mapping procedures, including the composite interval mapping (CIM) method, the mixed linear model (MLM-GWAS) method and the RTM-GWAS method, and the best mapping procedure was selected for the analysis of the 100-seed weight genetic system in NJRIKY population through comparing their detection power, including the detected number of QTLs and total phenotypic variation explained. 【Result】The 100-seed weight of Kefeng-1 and NN1138-2 were 9.0 g and 17.9 g, respectively, showing significant difference. The genotypic coefficient of variation and heritability of the trait were 12.4% and 85.4%, respectively. These results indicated that the population was suitable for genetic analysis of 100-seed weight trait. The RTM-GWAS procedure performed the best with the largest number of QTLs (57) explaining the most phenotypic variation (PVE=70.78%), while a total of 14 and 6 QTLs contributing 56.47% and 18.47% phenotypic variation were detected using CIM and MLM-GWAS, respectively. The 57 QTLs detected from the RTM-GWAS distributed on 19 chromosomes, of which 41 QTLs overlapped with 81 QTLs identified from 30 bi-parental populations in the literature. Furthermore, the PVE of 57 QTLs ranged from 0.03% to 7.57%, of which 16 QTLs were novel ones, including one large contribution major QTL Sw-09-2 (PVE>3%). Furthermore, 20 QTLs had significant interaction effect with environment. A total of 36 candidate genes were annotated from 37 QTLs through χ2 test between SNPLDB markers and SNPs harboring in the predicted genes, of which 4 candidate genes were from the large contribution QTLs and other 32 candidate genes were from the small contribution QTLs. These candidate genes were included in different biological processes, of which 13 candidate genes were grouped in seed development directly, and the remaining candidate genes were grouped into different functions, such as transport, transcriptional regulators, etc., indicating that these genes from different biological pathways regulate the expression of 100-seed weight trait in NJRIKY together. 【Conclusion】Among the three different mapping procedures, RTM-GWAS procedure is the most powerful one which can provide a relatively thorough detection of 100-seed weight QTLs in NJRIKY population, therefore, it is more suitable for QTL mapping in bi-parental population such as RIL population. The candidate genes with various functions jointly regulated the complex expression of 100-seed weight trait.