A joint learning approach for genomic prediction in polyploid grasses

Abstract

Poaceae, among the most abundant plant families, includes many economically important polyploid species, such as forage grasses and sugarcane (Saccharum spp.). These species have elevated genomic complexities and limited genetic resources, hindering the application of marker-assisted selection strategies. Currently, the most promising approach for increasing genetic gains in plant breeding is genomic selection. However, due to the polyploidy nature of these polyploid species, more accurate models for incorporating genomic selection into breeding schemes are needed. This study aims to develop a machine learning method by using a joint learning approach to predict complex traits from genotypic data. Biparental populations of sugarcane and two species of forage grasses (Urochloa decumbens, Megathyrsus maximus) were genotyped, and several quantitative traits were measured. High-quality markers were used to predict several traits in different cross-validation scenarios. By combining classification and regression strategies, we developed a predictive system with promising results. Compared with traditional genomic prediction methods, the proposed strategy achieved accuracy improvements exceeding 50%. Our results suggest that the developed methodology could be implemented in breeding programs, helping reduce breeding cycles and increase genetic gains.

Multiomic investigation of sugarcane mosaic virus resistance in sugarcane

“…Importantly, our results arose from a highly restricted SNP set obtained by FS, composed of only 73 markers, none of which had been identified by GWAS. A similar joint learning methodology that is based on FS and ML and combines classification and regression strategies has recently been shown to be highly suitable for the genomic prediction of several agronomic traits of sugarcane and polyploid forage grass species (Aono et al ., 2022).…”

A divide-and-conquer approach for genomic prediction in rubber tree using machine learning

“…For evaluating the genotypic profile of individuals in EG1 and EG2, we performed principal component analyses (PCAs) in R statistical software 58 with the ggplot2 package 71 . Additionally, for evaluating the overall correspondences between genotypic and phenotypic data, we colored the PCA scatter plots with the BLUPs estimated for SC trait, as performed by 72 .…”

Genome-wide association reveals novel insights into the molecular mechanisms regulating stem volume inPinus taeda

“…Contrasted with the predictions obtained with the models estimated using the entire set of markers, we evaluated the feasibility of feature selection (FS) techniques for subsetting the SNP data through putative phenotype-genotype associations. We selected the intersection between at least two out of three methods established (Aono et al, 2020; Aono et al, 2022a): (i) the gradient tree boosting (GTB) regressor model, (ii) Pearson correlations (maximum p value of 0.05), and (iii) the support vector machine (SVM) regression system. With such a dataset, we evaluated the importance of each SNP for prediction by calculating their feature importance using two different tree-based ML algorithms: decision tree (DT) and random forest (RF).…”

“…Considering our genotype dataset as a matrix , with n genotypes and m loci codified 𝑍 𝑛×𝑚 as 0 (reference homozygote), 1 (heterozygote) and 2 (alternative homozygote), the SNP effects ( Contrasted with the predictions obtained with the models estimated using the entire set of markers, we evaluated the feasibility of feature selection (FS) techniques for subsetting the SNP data through putative phenotype-genotype associations. We selected the intersection between at least two out of three methods established (Aono et al, 2020;Aono et al, 2022a): (i) the gradient tree boosting (GTB) regressor model, (ii) Pearson correlations (maximum p value of 0.05), and…”