Genomic Prediction yield of Oryza Sativa Using Machine Learning and Deep Learning

Abstract

Breeding value prediction plays a crucial role in improving crop breeding by accurately anticipating the genetic value of phenotypic traits. However, existing methods often lack accuracy in predicting genomic estimated breeding values (GEBVs) and do not sufficiently focus on regression-based approaches. To address this challenge, we propose a novel methodology for genomic prediction of phenotypic trait yield using a two-level classification approach.

In the first phase of our methodology, termed Genomic Prediction of Phenotypic Trait Yield using Two-Level Classification , we perform classification on the biological sequences of a subpopulation of Oryza sativa (rice). These sequences are then clustered based on the leaves of a phylogenetic tree, utilizing the Unweighted Pair Group Method with Arithmetic Mean (UPGMA) algorithm.

In the second phase, we employ machine learning techniques such as Multiple Linear Regression (MLR) to predict GEBVs and achieved remarkable accuracy ranging from 99 to 100 percent on the subpopulations of rice. By integrating the phylogenetic clustering approach and MLR-based prediction, our methodology demonstrates promising results for accurately predicting GEBVs, which can be passed on as genetic value to subsequent generations of offspring.

This research highlights the potential of our two-level classification approach in improving the accuracy of genomic prediction for phenotypic trait yield in rice breeding programs. The findings contribute to the development of enhanced breeding strategies, enabling more efficient selection of desired traits and facilitating the development of genetically improved crop varieties.