Identification of SNPs and genes related to grain yield in rice using Machine Learning

Abstract

The aim of this project is to provide a series of molecular markers and candidate genes, both associated with grain yield to assist in the development of new commercial rice strains and cultivars. For this, the project will use two databases from Embrapa's Nuclear Rice Collection, one phenotypic, generated from a series of nine field experiments where its 550 accessions were evaluated for productivity, and another genotypic, obtained by characterizing these accessions, where 445,589 SNPs were generated. Through the development of machine learning algorithms, productivity-associated SNPs will be identified, which, after validation, will enter the assisted selection routine of the rice breeding program. The team of this proposal (Embrapa and UFSCar) recently worked with the same strategy studying the amylose content (AC) character of the grain, and a SNP marker was found and validated which discriminates genotypes that have low TA from genotypes with high TA. This positive result motivated the team to elaborate the proposal for a highly relevant character of interest for the genetic improvement of rice from Brazil and the world. The combined efforts of the UFSCar and Embrapa research groups, which work in different but complementary areas, will synergistically achieve new and important results for the development of science and agriculture in the country. (AU)