Genomic prediction of hybrids and segregating populations of tropical maize by computational intelligence models

• Grant number: 22/14078-0
• Support Opportunities: Regular Research Grants
• Start date: September 01, 2023
• End date: August 31, 2026
• Field of knowledge: Biological Sciences - Genetics - Plant Genetics
• Principal Investigator: Michele Jorge Silva Siqueira
• Grantee: Michele Jorge Silva Siqueira
• Host Institution: Escola Superior de Agricultura Luiz de Queiroz (ESALQ). Universidade de São Paulo (USP). Piracicaba , SP, Brazil
• Associated researchers: Antonio Augusto Franco Garcia ; Jayme Garcia Arnal Barbedo

Abstract

It is noteworthy that in breeding programs it is essential to find a hybrid combination of twoparents with a high frequency of favorable and genetically complementary alleles, whose hybridpresents several loci in heterozygosity and manifests high hybrid vigor, preferably with an averagehigher than the best parent. In addition, prior knowledge of the potential of segregating populationscan lead to a significant reduction in costs and time, since materials with unfavorable genecombinations will be discarded early and breeders can focus efforts only on promising populations,aiming at plant selection that contain the desired gene combinations.In the present project, predictive procedures will be used to predict the potential of theparent populations of tropical maize with two approaches, the first, in the short term, based on the prediction of good single hybrids and the second focus on the identification of parents for the formation of a population base. The potentiality and variability generated by promising hybrid combinations will be considered, measured in possible inbred progenies (RILs), thus allowing hybrid combinations with low performance to be eliminated even before the crossing is carried out, causing greater efforts to be concentrated in only populations promising. Thus, the use of phenotypic and genotypic information will be considered for the establishment of predictive models of hybrids and populations based on computational intelligence and machine learning approaches. Such approaches, when associated with genetic diversity studiesand evaluation of the potential per se of the progenitor lines, demonstrate high potential forprediction analysis and are still little used. The use of predictive models with computationalintelligence and machine learning approaches will collaborate in the efficiency to obtain promisinghybrid combinations, and in the elimination of low performance combinations even before thecrossing is carried out, causing greater efforts to be concentrated only promising populations. (AU)