Exploring Deep Learning for Complex Trait Genomic Prediction in Polyploid Outcrossing Species

Zingaretti, Laura M.; Gezan, Salvador A.; Ferrão, Luís Felipe Ventorim; Osorio, Luis F.; Monfort, Amparo; Muñoz, Patricio; Whitaker, Vance M.; Pérez‐Enciso, Miguel

doi:10.3389/fpls.2020.00025

Cited by 105 publications

(136 citation statements)

References 85 publications

Supporting

Mentioning

118

Contrasting

Order By: Relevance

“…The authors also pointed out that the use of DL methods in GP of polyploid plants allows one to exploit its non-linearity, and it has less restrictive assumptions in comparison to traditional linear model-based methods. Moreover, polyploid plants might present higher degrees of complete and partial intra-locus interactions compared with diploid species (Zingaretti et al, 2020). These results are not in agreement with this study, in which the PA of DL did not evidence differences between the contributions of the additive or dominance effects (Table 4), since DL was the best method in the prediction (in terms of PA) of all traits.…”

Section: Discussioncontrasting

confidence: 91%

“…These results are not in agreement with this study, in which the PA of DL did not evidence differences between the contributions of the additive or dominance effects (Table 4), since DL was the best method in the prediction (in terms of PA) of all traits. These findings may be due to the fact that Bellot et al (2018) and Zingaretti et al (2020) used the Convolutional Neural Networks, whereas in the present study, the LSTM method was used; however, we emphasize that other studies must be performed to corroborate this argument. Interestingly, Alves et al (2020) compared the predictive performance of GBLUP with ANN method in simulated traits considering different levels of dominance effects.…”

Section: Discussionmentioning

confidence: 66%

“…Particularly, the results of this genomic prediction showed that DLs performed comparatively better as narrow-sense heritability decreased and the contribution of dominance increased. Similarly, Zingaretti et al (2020) implemented GP in polyploid outcrossing species (i.e., strawberry and blueberry) and found that DL did not show clear advantages over the linear models BL and BRR, except when the nonadditive effects (dominance or epistasis) were important. The authors also pointed out that the use of DL methods in GP of polyploid plants allows one to exploit its non-linearity, and it has less restrictive assumptions in comparison to traditional linear model-based methods.…”

Section: Discussionmentioning

confidence: 99%

“…Several studies have reported that the combination of hyperparameters critically influences the predictive performance of the DL model, emphasizing the need to carefully optimize hyperparameters in the ML architectures (Glória et al, 2016;Bellot et al, 2018;Pérez-Enciso and Zingaretti, 2019;Zingaretti et al, 2020). Particularly, Glória et al (2016) found that a simple architecture of BRNN outperformed other more complex architectures (by adding layers and/or more complex activation functions) in terms of PA.…”

Section: Discussionmentioning

confidence: 99%

“…Therefore, in these two important outcrossing species, we have assessed ML architectures, which were compared with GBLUP and different linear Bayesian regression models (Bayes A, Bayes B, Bayes Cπ, BRR, and BL) and the non-linear model RKHS. One of the main difficulties in implementing DL models is to find the best hyperparameter configuration, which requires time and some basic understanding of hyperparameters to optimize (Pérez-Enciso and Zingaretti, 2019;Zingaretti et al, 2020). In this regard, we evaluated the importance of mini-batch and activation function hyperparameters in terms of predictive ability (PA) and computational time required.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Genome-Wide Prediction of Complex Traits in Two Outcrossing Plant Species Through Deep Learning and Bayesian Regularized Neural Network

et al. 2020

View full text Add to dashboard Cite

Genomic selection models were investigated to predict several complex traits in breeding populations of Zea mays L. and Eucalyptus globulus Labill. For this, the following methods of Machine Learning (ML) were implemented: (i) Deep Learning (DL) and (ii) Bayesian Regularized Neural Network (BRNN) both in combination with different hyperparameters. These ML methods were also compared with Genomic Best Linear Unbiased Prediction (GBLUP) and different Bayesian regression models [Bayes A, Bayes B, Bayes Cπ, Bayesian Ridge Regression, Bayesian LASSO, and Reproducing Kernel Hilbert Space (RKHS)]. DL models, using Rectified Linear Units (as the activation function), had higher predictive ability values, which varied from 0.27 (pilodyn penetration of 6 years old eucalypt trees) to 0.78 (flowering-related traits of maize). Moreover, the larger mini-batch size (100%) had a significantly higher predictive ability for wood-related traits than the smaller mini-batch size (10%). On the other hand, in the BRNN method, the architectures of one and two layers that used only the pureline function showed better results of prediction, with values ranging from 0.21 (pilodyn penetration) to 0.71 (flowering traits). A significant increase in the prediction ability was observed for DL in comparison with other methods of genomic prediction (Bayesian alphabet models, GBLUP, RKHS, and BRNN). Another important finding was the usefulness of DL models (through an iterative algorithm) as an SNP detection strategy for genome-wide association studies. The results of this study confirm the importance of DL for genome-wide analyses and crop/tree improvement strategies, which holds promise for accelerating breeding progress.

show abstract

Section: Discussioncontrasting

confidence: 91%

Section: Discussionmentioning

confidence: 66%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Genome-Wide Prediction of Complex Traits in Two Outcrossing Plant Species Through Deep Learning and Bayesian Regularized Neural Network

et al. 2020

View full text Add to dashboard Cite

show abstract

Heterosis for biomass yield and other traits in ‘Alamo’ × ‘Kanlow’ switchgrass populations

et al. 2021

View full text Add to dashboard Cite

Heterosis plays an important role in switchgrass (Panicum virgatum L.) breeding. To identify superior clones for hybrid development through seasonal single cut, 10 crosses derived from ‘Alamo’ × ‘Kanlow’ and parents were evaluated. The experiment was planted in 2018 using a randomized complete block design with two replications at two locations in Tennessee: the East Tennessee Research and Education Center (ETREC), Knoxville, and the Plateau Research and Education Center (PREC), Crossville. The crosses and their parents were evaluated for biomass yield, plant height, and clonal mass in the fall of 2019 and 2020. The crosses were significantly different in biomass yield within and across locations (P ≤ .05). Average biomass yield of the crosses ranged from 7.1 to 8.7 Mg ha−1 at ETREC and from 10.0 to 11.8 Mg ha−1 at PREC. The crosses demonstrated an average of 46% mid‐parent heterosis (MPH) and 24% high‐parent heterosis (HPH) for biomass yield across the two locations. The MPH and HPH were 3 and −1.3% for plant height and 12 and 1% for clonal mass, respectively. Moderate to strong positive association was found between clonal mass scores and biomass yield at individual locations across years and across locations and years (r = 0.37–0.79, P ≤ .05). The association between biomass yield and plant height was inconsistent (r = −0.24 to 0.67, P ≤ .05). The findings of biomass yield heterosis in these crosses suggested the potential of hybrid cultivars for biomass yield improvement in lowland switchgrass.

show abstract

The application of pangenomics and machine learning in genomic selection in plants

et al. 2021

View full text Add to dashboard Cite

Genomic selection approaches have increased the speed of plant breeding, leading to growing crop yields over the last decade. However, climate change is impacting current and future yields, resulting in the need to further accelerate breeding efforts to cope with these changing conditions. Here we present approaches to accelerate plant breeding and incorporate nonadditive effects in genomic selection by applying state-of-the-art machine learning approaches. These approaches are made more powerful by the inclusion of pangenomes, which represent the entire genome content of a species. Understanding the strengths and limitations of machine learning methods, compared with more traditional genomic selection efforts, is paramount to the successful application of these methods in crop breeding. We describe examples of genomic selection and pangenome-based approaches in crop breeding, discuss machine learning-specific challenges, and highlight the potential for the application of machine learning in genomic selection. We believe that careful implementation of machine learning approaches will support crop improvement to help counter the adverse outcomes of climate change on crop production.

show abstract

Exploring Deep Learning for Complex Trait Genomic Prediction in Polyploid Outcrossing Species

Cited by 105 publications

References 85 publications

Genome-Wide Prediction of Complex Traits in Two Outcrossing Plant Species Through Deep Learning and Bayesian Regularized Neural Network

Genome-Wide Prediction of Complex Traits in Two Outcrossing Plant Species Through Deep Learning and Bayesian Regularized Neural Network

Heterosis for biomass yield and other traits in ‘Alamo’ × ‘Kanlow’ switchgrass populations

The application of pangenomics and machine learning in genomic selection in plants

Contact Info

Product

Resources

About