title | author | date | output | params | |
---|---|---|---|---|---|
Single Trial Analysis Report |
Contact:<a href = 'https://github.com/Breeding-Analytics/bioflow' target = '_blank'>Breeding Analytics Team, OneCGIAR</a> [email protected] |
January 16, 2025 |
html_document |
|
The objective of this dashboard is to help scientist to understand the following points:
-
Overall number of designations and environments included in the single trial analysis (input)
-
High-level summary statistics of the phenotypic information included (input)
-
Observed spatial variation in the different environments when coordinates of the field exist (input)
-
Genetic variance and other genetic parameters observed in the different environments for the different traits (output)
-
Individual adjusted means for each trait by environment combination (output)
-
Phenotypic correlation between environments for the traits present (output)
Understanding these data features should allow the scientist to identify trait by environments combinations that have enough genetic signal and take the decision of which to include in the multi-trial analysis. It should also allow the scientist to assess the quality of the trials conducted and take corrective measures (e.g., change service providers, improve practices, etc.).
No coordinates available. Skipping planting map.
The following table allows to see how many locations had data for the different traits. You may want to review if the phenotyping capacity can deal with the complexity of the trait (e.g., genotype by environment interaction) or if more resources should be deployed. Also you may want to check if you collected information from all the trials conducted.
The following table allows you to verify some quality metrics (KPIs) for the different trait by environment combinations.
Trait to filter:
This barplot allows you to see the variance components values and ratios for the trait by environment combinations and identify good quality trials.
Trait to filter:
The following barplot is designed to provide a high-level view of estimated parameters such as reliability, heritabiliy, coefficient of variation and others.
This section aims to provide the modeling table for the analysis in order to keep track of which environments were used in the analysis, what was the final model used for each trait and other potentially important parameters for future reference.
The adjusted means in the following visuualizations are the result of fitting a experimental-design agnostic mixed model where everything that can be fitted will be fitted in order to remove as much spatial noise as possible. That means that if a trial has block and incomplete block information both will be fitted. If the trial has also row and column information it will also be fitted together with a spatial kernel (Rodriguez-Alvarez et al., 2018). These table of adjusted means will be used as input information for the multi-trial analysis. We recommend you to don't take any selection decision at this point and wait until the multi-trial analysis is fitted.
The following table allows you to check the trait by environment adjusted means for the different individuals in wide format.
Environment to filter:
The following boxplot allows you to see the distribution of predicted values by trait (y-axis) in the different environments to double check that everything looks OK.
Trait to filter:
In the following plot you can observe the comparison between the top 30 designations from each entry type category for the different traits. If a category has less than a 30 designations all individuals are displayed. This should allow you to identify the top designations in each environment. We would NOT recommend you to use this for selection of parents or products. Wait until you have the results of the multi-trial analysis and selection indices.
The following plot aims to show the correlation between BLUEs or BLUPs (depending on the parameter settings) among the different environments for the traits available in order to identify if there is one or more environments that do not align with the target population of environments (i.e., negatively correlated with the main cluster across most environments). You may want to exclude such environments from the multi-trial analysis (MTA) to ensure that selected designations in the MTA achieve genetic gain in the main cluster of environments.
Trait to filter:
Velazco, J. G., Rodriguez-Alvarez, M. X., Boer, M. P., Jordan, D. R., Eilers, P. H., Malosetti, M., & Van Eeuwijk, F. A. (2017). Modelling spatial trends in sorghum breeding field trials using a two-dimensional P-spline mixed model. Theoretical and Applied Genetics, 130, 1375-1392.
Rodriguez-Alvarez, M. X., Boer, M. P., van Eeuwijk, F. A., & Eilers, P. H. (2018). Correcting for spatial heterogeneity in plant breeding experiments with P-splines. Spatial Statistics, 23, 52-71.
R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.
Boer M, van Rossum B (2022). LMMsolver: Linear Mixed Model Solver. R package version 1.0.4.9000.