Genomic Selection Based Genome Optimization Model For Design Breeding In Maize

Genomic selection（GS）is a promising strategy for selective breeding through predicting phenotypes of individuals from their genotypes,which can effectively reduce the high expense of field phenotyping for phenotypic selection（PS）.The employment of doubled haploid（DH）technology in maize has vastly accelerated the efficiency of developing inbred lines and GS is preferable for selection of superior lines.The application of GS in plant breeding practices still faces many challenges,thus continuous innovation of breeding frameworks,methods and toolkits are required to serve as a strong support for maize breeding even crop breeding.Here,two breeding-assisted and decision-making toolkits were developed.To prove the utilization and features of them,a maize population and partial F₁ generations were used for testing.And the genome optimization model was applied to design the breeding process and improved the genetic gain.We provide a new perspective and two analytical tools for assisting breeding decision-making driven by Big Data,which will accelerate breeding process and facilitate maize breeding step into Breeding 4.0 era.The development of an integrated toolkit for genomic selection.The implementation of GS encounters several computational challenges,including the construction of complex GS models,the optimization of multifarious GS models,and the choice of appropriate GS models.Here,an integrated R toolkit named G2P（genotype-to-phenotype）was developed to facilitate genotype-to-phenotype prediction for GS and related analysis.G2P accepts multiple formats of input genotypic and provides a comprehensive set of analysis functions for quality control and data pre-processing,including filtration,imputation,and conversion of genotypic coding.G2P also presents a unified framework that incorporates 16 state-of-the-art GS methods and13 evaluation metrics,thereby decreasing burden of coding and increasing flexibility for model construction,optimization,evaluation,and recommendation.Additionally,G2P offers two ensemble strategies to consolidate multiple predictions resulting from different GS models for enhancing the prediction performance from genotypes to phenotypes.In order to balance flexibility and simplicity of G2P,on the basis of modularization,all core functions are stream-lined and reduce to an ultimate report in HTML format with interactive feature.G2P has been packed into a Docker container,ensuring the ease of use,reliability,reproducibility,and rep-licability for crop GS analysis.A maize inbred population was used to illustrate the computa-tional effectiveness and features of G2P in GS relative analysis.Genome optimization design model.The decision-making of GS-assisted breeding usu-ally relies on prediction results directly that may lose genetic diversity and cannot guide the next round improvement of lines.Genome optimization via virtual simulation（GOVS）,sim-ulates a virtual genome by algorithm.This virtual genome is supposed to encompass the most abundant“optimal genotypes”or“advantageous alleles”for target trait in the genetic pool.Nowadays,such a virtually optimized genome may never be developed in reality,but plot an optimal route to direct breeding decision.GOVS assists in the selection of superior lines based on the genomic fragments that a line contributes to the virtual genome.The assumption is that the more fragments of optimal genotypes a line used to assemble virtual genome,the higher likelihood of the line favored in corresponding phenotype.Additionally,the selected lines by GOVS are complementary to each other of advantageous alleles contributed to virtual genome.This feature can direct the next round population development,or facilitate plotting the opti-mal route for DH production,whereby the fewest lines and F₁ combinations are needed to pyramid a maximum number of advantageous alleles in the new DH lines.GOVS,an R pack-age,was developed to achieve the conception of genome optimization.In this work,we im-plemented GOVS using the genotype and phenotype data of a maize population and its F₁progeny.The precision of lines selection by GS and GOVS was compared,and GOVS showed competitive performance versus GS.It’s worth noting that GOVS is not the substitution but the complement of GS,providing new perspective and option.Genomic optimization design accelerates maize breeding.Based on genome optimiza-tion model,a maize breeding framework was designed.A maize inbred population and their F₁ progeny cross with ZHENG58 were used to implement GOVS for the assembly of virtually optimized genome of ear weight（EW）.As a result,253 F₁ hybrids were selected under the consideration of their contribution to virtually optimized genome,indicating that the 253 cor-responding maternal lines have potential to obtain favorable hybrids by crossing with superi-orly paternal testers.Then,three superiorly paternal testers were selected from 30 paternal testers by the analysis of general combining ability（GCA）to generate 80 hybrid combinations.The 80 selected combinations are untested in field but are considered have potential surpassing the check.Finally,the results of field phenotyping for 80 combinations showed that GOVS accelerated the progress of genetic gain and reduced the cost in filed.In summary,DH technology has been widely applied in maize breeding industry,as it greatly shortens the period of developing homozygous inbred lines via bypassing several rounds of self-crossing.The current challenge is how to efficiently screen the large volume of inbred lines based on genotypes.Simultaneously,the advantageous alleles for target trait are required to pyramid rapidly with purpose.Here,an integrated toolkit for genomic selection named G2P and another toolbox GOVS for genome optimization were developed,GOVS complements the traditional genomic selection model.A maize inbred population and their F₁progeny were used to illustrate the implements of G2P and GOVS.The incorporation of DH production,GS and genome optimization will ultimately improve genomically designed breeding in maize driven by Big Data,whereby accelerating the maize breeding.