Genome wide selection in sugarcane using machine learning and complex networks for economically important traits

Abstract

The use of molecular markers SNPs (Single Nucleotide Polymorphisms), although made a major contribution to genetic improvement programs, has limited application in the case of sugarcane, a species with high genomic complexity. Due to the lack of biocomputational tools compatible with the uniqueness of the species, specific methodologies have to be used. With immeasurable importance to the sectors of biofuels and sugar manufacturing, the improvement of the species and the cultivars with characters of economic importance represent a major contribution to the world economy. Due to the polygenic nature of quantitative phenotypic characteristics and the consequent complexity of application of methods for genotype-phenotype association, genomic selection methodologies have been shown to be an alternative to traditional approaches for marker-assisted selection. The present work proposes an innovative approach for realization of selection in cane. Using an array of putative SNPs identified with the use of sequencing genotyping data and software adapted to polyploidy and aneuploidy of the species, an association modeling will be used between these markers using complex networks. By means of detection methods of communities in the built network, it is intended to carry out data imputation, dimensionality and identification of SNPs with related biological categories. For constructionof the selection model will be used machine learning techniques as models nonlinear, constructed to capture the specificity of the cane genome. (AU)