Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020 Jul 7:11:1001.
doi: 10.3389/fpls.2020.01001. eCollection 2020.

Genomic Prediction of Agronomic Traits in Common Bean (Phaseolus vulgaris L.) Under Environmental Stress

Beat Keller  1   2 , Daniel Ariza-Suarez  1 , Juan de la Hoz  1 , Johan Steven Aparicio  1 , Ana Elisabeth Portilla-Benavides  1 , Hector Fabio Buendia  1 , Victor Manuel Mayor  1 , Bruno Studer  2 , Bodo Raatz  1
Affiliations

Genomic Prediction of Agronomic Traits in Common Bean (Phaseolus vulgaris L.) Under Environmental Stress

Beat Keller et al. Front Plant Sci. .

Abstract

In plant and animal breeding, genomic prediction models are established to select new lines based on genomic data, without the need for laborious phenotyping. Prediction models can be trained on recent or historic phenotypic data and increasingly available genotypic data. This enables the adoption of genomic selection also in under-used legume crops such as common bean. Beans are an important staple food in the tropics and mainly grown by smallholders under limiting environmental conditions such as drought or low soil fertility. Therefore, genotype-by-environment interactions (G × E) are an important consideration when developing new bean varieties. However, G × E are often not considered in genomic prediction models nor are these models implemented in current bean breeding programs. Here we show the prediction abilities of four agronomic traits in common bean under various environmental stresses based on twelve field trials. The dataset includes 481 elite breeding lines characterized by 5,820 SNP markers. Prediction abilities over all twelve trials ranged between 0.6 and 0.8 for yield and days to maturity, respectively, predicting new lines into new seasons. In all four evaluated traits, the prediction abilities reached about 50-80% of the maximum accuracies given by phenotypic correlations and heritability. Predictions under drought and low phosphorus stress were up to 10 and 20% improved when G × E were included in the model, respectively. Our results demonstrate the potential of genomic selection to increase the genetic gain in common bean breeding. Prediction abilities improved when more phenotypic data was available and G × E could be accounted for. Furthermore, the developed models allowed us to predict genotypic performance under different environmental stresses. This will be a key factor in the development of common bean varieties adapted to future challenging conditions.

Keywords: common bean (Phaseolus vulgaris L.); drought; genome-wide association studies (GWAS); genomic selection; genotype × environment interactions; low phosphorus stress; plant breeding.

PubMed Disclaimer

Figures

Figure 1
Figure 1
Phenotypic variability of 100 seed weight (100SdW), days to flowering (DF), days to physiological maturity (DPM), and seed yield evaluated in up to twelve field trials. The trials were carried out between 2013, first planting season ‘A’ and 2018 third season ‘C’ under drought, irrigated and different phosphorus (P) conditions in Darien (Dar) and Palmira (Pal), Colombia. In total, 481 common bean lines were evaluated (between 156 and 345 lines per trial). Best linear unbiased estimators (BLUEs) were obtained for each trait and trial, adjusting for spatial effects in the field in a first-stage analysis. In addition, weighted BLUEs were calculated over all trials in a second-stage analysis.
Figure 2
Figure 2
Assessment of population structure for the VEF population including 481 common bean lines with 5,820 SNP markers: (A) The principal component analysis shows the location of each genotype defined by the eigenvectors of the first and second principal components. (B) The unrooted neighbor-joining tree indicates the absence of a clear differentiation pattern between lines. The length of the lines in the tree shows the simple matching distance.
Figure 3
Figure 3
Genome-wide association study for in total 481 common bean lines using second-stage best linear unbiased estimators (BLUEs) across twelve trials for the traits 100 seed weight (100SdW), days to flowering (DF), days to physiological maturity (DPM), and seed yield. (A) The Manhattan plots show the significance of associations of every SNP marker for each trait calculated using the BLINK algorithm, implemented in GAPIT. The vertical line indicates a common QTL for DF, DPM and seed yield. The horizontal line shows the false discovery rate (p = 0.05) to identify significant associations. (B) Q–Q plots show the distribution of the negative logarithm of expected and observed p values.
Figure 4
Figure 4
Genomic prediction abilities of seed yield in response to the number of utilized markers evaluated in the VEF and the MAGIC populations. The markers used for prediction were chosen either randomly (the white line and its gray stripe show the average prediction ability and its corresponding standard deviation) or based on LD and MAF parameters (colored ranges, the middle point and its error bar represent the average prediction ability and its standard deviation). The distribution of values in this plot corresponds to 100-fold cross validations with 70:30 training:validation population partitioning.
Figure 5
Figure 5
Genomic prediction abilities evaluated separately for each of the twelve trials as well as over all seasons in a total of 481 lines. The 100 seed weight (100SdW), days to flowering (DF), days to physiological maturity (DPM), and seed yield were evaluated between 2013 and 2018 in season ‘A’ or ‘C’, under drought, irrigated and different soil phosphorus (P) conditions in Darien (Dar) and Palmira (Pal), Colombia. The model predictions were based on 4,962 SNP markers using Bayesian ridge regression. In the overall predictions, new lines were predicted based on the second-stage best linear unbiased estimators (BLUEs) obtained from all the trials. Black dots and triangles indicate broad-sense and genomic heritability, respectively.
Figure 6
Figure 6
Genomic prediction abilities for new lines in new seasons compared for different models using 100 fold cross validation. Different genotype × environment interactions were considered to improve the basic SNP marker model among different traits: Modeling the effects of drought and irrigated conditions in Palmira and the location effect of Palmira separately (Marker * Env model), the stress effect for drought and low P conditions (Marker * Stress model) or the Marker * Env with fixed QTL effects (QTL model). The 100 seed weight (100SdW), days to flowering (DF), days to physiological maturity (DPM), and seed yield were evaluated in up to twelve trials. The trials were carried out between 2013 first planting season ‘A’ and 2018 third season ‘C’ under drought, irrigated and different phosphorus (P) conditions in Darien (Dar) and Palmira (Pal), Colombia.
Figure 7
Figure 7
Genomic prediction abilities for seed yield in a breeding program using chronological data accumulated up to the predicted season (colored boxplots), and data over all seasons (black boxplots) to predict new lines in new seasons using 100 fold cross validation. In total 4,962 markers were fitted using Bayesian ridge regression and the Marker * Env model (see Material and Methods). In total, twelve trials between 2013 first planting season ‘A’ and 2018 third season ‘C’ under drought, irrigated and different phosphorus conditions in Darien (Dar) and Palmira (Pal), Colombia were evaluated.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Achard S. (2012). brainwaver: Basic wavelet analysis of multivariate time series with a visualisation and parametrisation using graph theory. R Project for Statistical Computing (R-Packages).
    1. Acosta-Pech R., Crossa J., de los Campos G., Teyssèdre S., Claustres B., Pérez-Elizalde S., et al. (2017). Genomic models with genotype × environment interaction for predicting hybrid performance: an application in maize hybrids. Theor. Appl. Genet. 130, 1431–1440. 10.1007/s00122-017-2898-0 - DOI - PubMed
    1. Allier A., Moreau L., Charcosset A., Teyssèdre S., Lehermeier C. (2019). Usefulness Criterion and Post-selection Parental Contributions in Multi-parental Crosses: Application to Polygenic Trait Introgression. G3: Genes Genomes Genet. 9, 1469–1479. 10.1534/g3.119.400129 - DOI - PMC - PubMed
    1. Asfaw A., Blair M. W., Struik P. C. (2012). Multienvironment Quantitative Trait Loci Analysis for Photosynthate Acquisition, Accumulation, and Remobilization Traits in Common Bean Under Drought Stress. G3: Genes Genomes Genet. 2, 579–595. 10.1534/g3.112.002303 - DOI - PMC - PubMed
    1. Assefa T., Assibi Mahama A., Brown A. V., Cannon E. K. S., Rubyogo J. C., Rao I. M., et al. (2019). A review of breeding objectives, genomic resources, and marker-assisted methods in common bean (Phaseolus vulgaris L.). Mol. Breed. 39, 20. 10.1007/s11032-018-0920-0 - DOI

LinkOut - more resources