Development And Application Of The Genome-wide Selection Platform CropGS-Hu

Genomic Selection(GS)stands as a widely used breeding technology,which constructs GS models from genetic marker information spanning the entire genome to predict trait phenotypes and select individuals with desired traits in earlier generations.In order to maximize the utilization of germplasm resources,advance crop breeding,integrate genomic and phenotypic data resources of crop populations as well as establish open breeding databases and functional platforms,a crop genome selection platform called Crop GS-Hub(https://iagr.genomics.cn/Crop GS/)has been developed in this study.Genotypes,agronomic phenotypes,and QTLs were systematically gathered from 14 populations across seven pivotal crops,namely rice,maize,soybean,cotton,foxtail millet,chickpea,and rapeseed.Employing a unified and standardized genome-wide association study(GWAS)pipeline and significance threshold determination approach,we obtained the association variants for each agronomic trait within each species.To enhance accessibility and visualization of variation information and GWAS results,the Crop GS-Hub platform offers a genome browser tool.Furthermore,we integrated diverse genome selection modeling algorithms based on both traditional statistics and machine learning approaches,and developed a"one-stop"analysis system from sequencing reads to the generation of a phenotypic prediction report,which is of great significance to the researches in genomic selection and practical applications in crop breeding.The main research results are as follows:1)Fourteen high-quality crop populations with genotypes and phenotypes were collected:covering seven major crops including maize(Zea mays),rice(Oryza sativa),cotton(Gossypium hirsutum),foxtail millet(Setaria italica),chickpea(Cicer arietinum),rapeseed(Brassica napus)and soybean(Glycine max).The datasets underwent uniform cleaning and filtering at the sample,genotype,and phenotype levels in this study,resulting in a total of approximately 30 thousand crop samples,involving349 important agronomic traits.2)Relevant GWAS resources and variant annotations were acquired:Using a unified and standardized GWAS analysis method along with significance threshold criteria,we extracted information regarding genetic variants associated with each agronomic trait from the above dataset.A total of 16,641 significant single nucleotide polymorphism(SNP)association sites were identified,and detailed annotation of variant and phenotypic effects was carried out for these associated SNPs.3)GS models for related agronomic traits were constructed:We evaluated the effect of SNP type on the prediction accuracy of the GS models,and determined the input criteria for GS model construction using LD-pruned(r~2<0.5)synonymous SNPs.Finally,2,094 genomic selection models were constructed for phenotypic prediction and breeding selection using 6 algorithms:GBLUP,RRBLUP,Bayes Cpi,Bayes L,Bayes R and Light GBM,and their predictive utility was verified.4)Optimization of genotyping process and development of a rapid genotyping toolkit,Single Nucleotide Polymorphism Genotyping Toolkit(SNPGT):It is capable of rapidly genotyping at thousands or even tens of thousands of target loci in a population for SNPGT and can be run in a local environment,both on Linux and Windows.5)The genome-wide selection platform,Crop GS-Hub,was developed:The platform can be divided into two main components:data resources and GS function modules.In the data resources section,the platform integrates the population data and GWAS resources of the seven crops mentioned earlier,offering a genome browser tool for convenient visualization of variant information and GWAS results.On the other hand,the GS function module incorporates six GS algorithms,constituting a comprehensive"one-stop"analysis system that covers the entire process from inputting sequencing data,constructing GS models,to generating phenotype prediction reports.