| The high-throughput phenotyping platform with multiple sensors can non-destructively and efficiently measure plant type and physiological traits,providing an effective method for collecting high-quality phenotype data.In order to explore the genetic constitution of soybean traits measured with high-throughput spectral and their effectiveness in biomass prediction,based on the high-throughput phenotyping platform of Nanjing Agricultural University,plant type and physiological traits of 1,234 soybean germplasm accessions were measured every 3days from 6th to 30th days after sowing,resulting in 9 different time points(DAS6,DAS9,DAS12,...,DAS30).Firstly,genome-wide association studies were performed for 10 high-throughput traits(HEI,SUR,WID,MAR,MBR,OBS,CHL,NDVI,F0,Fm)based on 606soybean accessions genotyped with genome-wide SNP markers.Secondly,420 soybean accessions were used to establish soybean aboveground biomass prediction model.This study demonstrated that the high-throughput platform is of important significance for soybean germplasm innovation.The main results are as follows:(1)Phenotypic variation of plant type and physiological traits at soybean seedling stage based on high-throughput measurement.Among the six plant type traits,HEI,SUR and WID reflect the side plant type traits of soybean,and had the highest heritability(80.97%,87.33%and 87.28%)at the 12th,18th and 12th day after sowing.The HEI,SUR and WID ranged from 148.73-418.20(pixels,the same below),1.71-15351.71 and 17.25-169.10,with averages of 280.19,6174.75 and 90.19,respectively.MAR,MBR and OBS reflect the canopy plant type traits of soybean,and had the highest heritability(77.29%,87.04%and 89.01%)at the 24th,15th and 18th day after sowing.The MAR,MBR and OBS ranged from 0.09-0.99,4.28-41.75 and 799.91-100204.59,with averages of 0.65,21.15 and 41723.84,respectively.CHL and NDVI reflect vegetation indexes physiological traits of soybean,and had the highest heritability(72.61%and 81.73%)at the 24th and 18th day after sowing.The CHL and NDVI ranged from 2.29-4.97 and 0.04-0.16,with averages of 3.65 and 0.10,respectively.The two chlorophyll fluorescence physiological traits F0 and Fm had the highest heritability(73.09%and 80.27%)at the 21th and 24th day after sowing.The F0 and Fm ranged from 556.14-1846.14 and 3041.32-6943.32,with averages of 1111.82 and 5138.21,respectively.(2)The aboveground biomass prediction model based on the phenotypic data measured by high-throughput at soybean seedling stage.With the phenotypic data measured by high-throughput at soybean seedling stage as independent variables and the above-ground biomass data manually measured as dependent variables,the KNN,LR,SVM and RF model were used to construct soybean above-ground biomass prediction models,respectively.The results showed that the R~2 and r of the RF model at 7 time points of DAS6-DAS24 are the highest among the four models(R~2:0.442-0.848,r:0.662-0.922),and the RMSE is the lowest(RMSE:2.341-1.258).The R~2(0.837,0.804)and r(0.918,0.902)of the RF model at the two time points of DAS27 and DAS30 are only lower than LR Model(R~2:0.844,0.812;r:0.920,0.903).Based on the RF model,the fresh weight prediction model constructed with the phenotypic data collected at DAS27 was the best,the R~2of the model was 0.926 and the RMSE was 1.466,and the r between the predicted and observed values was 0.963.The dry weight prediction model constructed with the phenotypic data collected at DAS24 was the best,the R~2 of the model was 0.948 and the RMSE was 0.301,and the r was 0.975.Determined that the phenotypic data collected 18-30 days after sowing of soybeans was used to construct the model with the best fit;verified that the constructed model has good application value in different times and different environments.(3)GWAS of plant type and physiological traits in soybean seedling stage.Using the RTM-GWAS(restricted two-stage multi-locus genome-wide association analysis)method,34,43 and 35 QTLs with 114,114 and 109 alleles were detected for HEI,SUR and WID,respectively,which explained 44.69%,46.73%and 43.96%of the phenotypic variation and candidate gene analysis annotated to 32,52 and 37 candidate genes for HEI,SUR and WID,respectively.A total of 31,41 and 47 QTLs with 106,144 and 136 alleles were detected for MAR,MBR and OBS,respectively,which explained 42.25%,52.39%,and 48.51%of the phenotypic variation and candidate gene analysis annotated to 26,39 and 35 candidate genes for MAR,MBR and OBS,respectively.A total of 33 and 33 QTLs with 104 and 107alleles were detected for CHL and NDVI,respectively,which explained 40.56%and 41.67%of the phenotypic variation and candidate gene analysis annotated to 38 and 27 candidate genes for CHL and NDVI,respectively.A total of 34 and 39 QTLs with 111 and 137 alleles were detected for F0 and Fm,respectively,which explained 42.74%and 47.97%of the phenotypic variation and candidate gene analysis annotated to 34 and 28 candidate genes for F0 and Fm,respectively.(4)Dynamic QTL analysis of soybean SUR at seeding stage.Descriptive statistics showed that the heritability of SUR ranged from 70.79%-87.33%and a total of 338 QTL(including 1005 alleles)were detected for SUR at 9 time points.Three QTLs could be detected at 4 different time points and the genetic contribution of a QTL(R~2)ranged from0.61%-4.11%.Nine QTLs could be detected at 3 different time points and the R~2 ranged from0.46%-3.83%.Thirty QTLs could be detected at two different time points and the R~2ranged from 0.37%-4.2%.A total of 239 QTLs could be detected only at one time point.The QTL detected at two or more time points were used to draw a QTL time heat map to showing in detail the dynamic changes of QTL R~2,p-value,and the number of alleles at different time points. |
PDF Full Text Request