Sparse kernel models provide optimization of training set design for genomic prediction in multiyear wheat breeding data

Abstract

The success of genomic selection (GS) in breeding schemes relies on its ability to provide accurate predictions of unobserved lines at early stages. Multigeneration data provides opportunities to increase the training data size and thus, the likelihood of extracting useful information from ancestors to improve prediction accuracy. The genomic best linear unbiased predictions (GBLUPs) are performed by borrowing information through kinship relationships between individuals. Multigeneration data usually becomes heterogeneous with complex family relationship patterns that are increasingly entangled with each generation. Under these conditions, historical data may not be optimal for model training as the accuracy could be compromised. The sparse selection index (SSI) is a method for training set (TRN) optimization, in which training individuals provide predictions to some but not all predicted subjects. We added an additional trimming process to the original SSI (trimmed SSI) to remove less important training individuals for prediction. Using a large multigeneration (8 yr) wheat (Triticum aestivum L.) grain yield dataset (n = 68,836), we found increases in accuracy as more years are included in the TRN, with improvements of ∼0.05 in the GBLUP accuracy when using 5 yr of historical data relative to when using only 1 yr. The SSI method showed a small gain over the GBLUP accuracy but

Genome-wide association and genomic prediction for iron and zinc concentration and iron bioavailability in a collection of yellow dry beans

“…Although the YBC represents a diverse population that might benefit from SSI, its training set consists of fewer than 185 individuals. This is in contrast to the thousands of individuals used in the training models by Lopez-Cruz and de los Campos (2021 ) and Lopez-Cruz et al (2022 ).…”

“…In a multi-generational wheat dataset spanning 8 years, the SSI model demonstrated superior prediction accuracy compared to GBLUP ( Lopez-Cruz et al, 2022 ). As Lopez-Cruz and de los Campos ( Lopez-Cruz and de los Campos, 2021 ) highlighted, SSI tends to achieve higher accuracies than GBLUP, especially in larger datasets.…”

“…The better performance of SSI compared to KA in seed yield could be the result of differences in allele frequencies and LD patterns in the families to be predicted compared to the YBC. Differences in allele frequencies and LD may lead to suboptimal estimation of breeding values using all sets of individuals to train the model ( Lopez-Cruz et al, 2022 ). Overall, prediction accuracies were slightly lower when non-Andean accessions were excluded from the training dataset.…”

“…The composition of the training data set and its relationship with the new lines to be predicted are crucial for maximizing prediction accuracy ( Isidro et al, 2021 ; de los Campos et al, 2013 ). While several approaches have been proposed to optimize the training population, most of them have the assumption that one training is optimal for all individuals in the testing dataset ( Lopez-Cruz et al, 2022 ). Considering that a high level of genetic heterogeneity is plausible in breeding programs, using only one optimal population may include individuals distantly related to the individuals in the testing dataset, reducing predictive ability ( Lorenz and Smith, 2015 ).…”

“…Recently, a sparce selection index (SSI) was proposed to identify a training set for each individual in the testing set ( Lopez-Cruz and de los Campos, 2021 ). The use of this approach has increased prediction ability in multigeneration data up to 10% and 17% in wheat and maize, respectively, compared to the GBLUP ( Lopez-Cruz et al, 2021 ; Lopez-Cruz et al, 2022 ).…”