Near-infrared Modeling And GWAS Analysis Of Cottonseed Components

Cotton（Gossypium spp.）is one of the important economic crops in the world.Cotton fiber is the main raw material of the textile industry,and cottonseed is an important by-product of the cotton industry.Cottonseed is rich in fatty acids and proteins and is one of the most important vegetable oil sources in the world.Cottonseed is an important source of protein for human and livestock,and can also be used as renewable raw materials for various industrial products,such as biofuels,lubricants and hydraulic oils.In the process of cotton breeding and production,it is often necessary to quickly and accurately analyze the protein content and fatty acid content of a large number of cottonseed samples.The purpose of this study was to establish prediction models for total protein（PR）content,seed total fatty acid（STA）content,cottonseed kernel total fatty acid（KTA）content,and the composition of major saturated fatty acids（myristic acid C14:0,palmitic acid C16:0,and stearic acid C18:0）with high accuracy and wide applicability by NIRS.The content of cottonseed components was estimated by the established near-infrared model,and the genome-wide association analysis was performed.The main results were listed as follows:1.In this experiment,135 Upland Cotton Germplasms from home and abroad were planted in two places for three years,and the original spectral information was obtained by scanning the shell-intact cottonseed with DA7200.The wavelength range used is 950nm-1650nm.Before the model construction,the standard normal variate（SNV）and the Savitzky–Golay derivative（first derivative）were used for raw spectra preprocessing.The protein,oil and five fatty acids contents of shell-intact cottonseed were determined by the reference method,respectively.The relative values of PR,KTA,STA,and five fatty acid content range from 0.541%to 61.848%,and the coefficient of variation ranges from 2.221%to 19.076%,which indicates that the above components have a wider variation in cottonseed,and are conducive to the construction of near-infrared calibration models.At the same time,PR was significantly negatively correlated with KTA,STA,MA,and SA.2.Near-infrared reflectance spectroscopy（NIRS）calibration equations using partial least squares regression for protein concentration,oil concentration,and five fatty acids of shell-intact cottonseeds were established based on 90 varieties,and the prediction abilities and regression effect of the calibration models were verified using 45 other varieties.Each equation was assessed based on the ratio of performance to deviation（RPD_p）.The prediction abilities of the NIRS calibration equations were basically consistent with external validation results.The coefficient of determination(R²_val),residual predictive deviation（RPD_p）were 0.926 and 3.687for protein,0.920 and 3.530 for STA,the results indicated that the calibration models of protein content and STA content were achieved with good accuracy and robustness(R²_val>0.900,RPD_p>3.500).The R²_valal values for KTA,linoleic acid（LNA）,stearic acid（SA）,myristic acid（MA）,and palmitic acid（PAM）were 0.878,0.876,0.863,0.860 and 0.841,respectively,and the RPD_p values were 2.866,2.836,2.697,2.676 and 2.506,respectively.The results showed that the calibration models of KTA,LNA,SA,MA and PAM were achieved with relatively high accuracy and relatively good stability(R²_val>0.840,RPD_p>2.500).The R²_valal values for OLA were 0.736,and the RPD_p values were 1.945.The results showed that the calibration models of OLA were achieved with bad accuracy and stability(R²_val<0.800,RPD_p<2.000).Therefore,the near-infrared spectroscopy can replace the conventional chemical analysis methods and can be used for rapid and nondestructive analysis of PR,KTA,STA,LNA,PAM,SA and MA in the shell-intact cottonseed.3.The chemical measurements and near-infrared predictions of eight cottonseed nutritional quality traits were used to correlate with SNP typing data（41832 polymorphic SNP markers）,respectively.Using a mixed linear model MLM（PCA+K）,20 SNP markers that were significantly associated with the nutritional quality of cottonseed were detected simultaneously in chemical measurements and near-infrared predictions.The SNP markers associated with PR,KTA,LNA,OLA,MA,PAM,STA and SA were 2,4,6,2,1,1,1,1,and3,respectively.They accounted for 66.67%,80%,100%,50%,50%,33.33%,33.33%and 75%of the number of SNP markers detected by chemical measurements,respectively.Among the co-detected SNP markers,the most significant SNP markers associated with PR,KTA,LNA,OLA,MA,and SA were consistent.The 20 co-detected SNP markers were mapped to 12 QTL regions.For PR,KTA,LNA,Ola,Ma,PAM,STA and SA,they were located in 2,3,2,1,1,1,1,1 and 1 QTL regions respectively,accounting for 66.67%,75%,100%,33.33%,100%,33.33%,33.33%,33.33%and 50%of the QTL regions determined by chemical measurements.Of the 12 QTLs detected by both methods,four QTLs（qGhLNA-c10,qGhKTA-c13,qGhPR-c13-2,and qGhSA-c24）have been reported in previous studies.83 candidate genes were initially obtained from 21 QTL regions screened by chemical measurements,including 24predominantly or specifically expressed in ovules.Three genes that may be related to cottonseed fatty acid synthesis were initially identified through functional annotation,Which are Gh_A07G2087,Gh_A07G2084,and Gh_A13G0396.The above results show that the calibration models of PR,KTA,LNA,MA and SA are of high accuracy and good stability,and the near-infrared prediction results can be used for genome-wide association analysis.