Skip to content
Envirotyping: shaping the breeding landscape
Environmental factors play a crucial role in plant breeding, as they influence the expression of desirable traits. In this context, envirotyping refers to characterizing and quantifying the environmental factors, such as temperature, soil quality and water availability, that influence the development and expression of an organism’s phenotype, such as growth rate, disease resistance, and especially yield.
Overall, this field of study is critical for understanding why phenotypes are what they are. For instance, it is possible to identify the main environmental components that drive phenotypic performance, understand at a deeper level how they affect the traits, explore the underpinning genotype by environment interactions, and even build prediction models for unseen environments.
Envirotyping’s depth: deciphering genetic-environmental interactions
To exemplify its use, I will present a case study on how the Louisiana State University (LSU) rice breeding program has used this information to optimize multi-environment trials (MET). The need for optimization is because the phenotypic evaluation is one of the most expensive parts of a breeding program. Consequently, identifying the best balance between the allocation of trials and the heritability (measurement of selection certainty) is one of the main questions to be answered.
For that, ten years of weather data during the rice growing season (temperature at two meters, precipitation, wind speed at two meters, relative humidity at two meters, and all-sky insolation incident on a horizontal surface) were retrieved from the NASA servers using the EnvRtype R package. However, the sky is not the limit! This dataset was further enriched with the addition of 13 soil characteristics (six physical and seven chemical), obtained using the SoilType R package. Finally, the most relevant weather and soil covariates were identified based on a feature selection algorithm. The criteria was to select a set of covariates that best explained two years of yield data from 25 genotypes across 15 locations (Figure 1).
Figure 1. The most important environment covariates (x-axis) to explain rice grain yield in several locations in the USA (y-axis). More details about the environmental covariates can be found in the envRtype and SoilType documentation.
Key covariates: unraveling genotype-environment connections
Using a few environmental-derived covariates (sand, clay, temperature range, and wind speed) in early crop stages, it was possible to explain 68% of the genotype by environment G×E interaction. Furthermore, using Gaussian kernels, it was possible to identify the relationship between environments (ERM, Figure 2), group the locations via a k-means clustering approach, and optimize the MET via mixed model equations (Figure 3).
Figure 2. Environment relationship matrix (ERM) between rice trials, USA.
Figure 3. Clustering of rice trials, USA. The clusters represent locations with similar environmental responses to rice yield.
Optimized MET: maximizing efficiency and accuracy
In the optimization step, a single location per cluster was identified to compose an optimized MET set (OPT_MET) based on the locations that maximized the heritability of the joint analysis. Hence, it was possible to reduce the total number of trials, better allocate the resources, and define the optimal MET in advance without significant losses in accuracy (heritability) compared to the original scenario (Full_MET) (Figure 4A). Furthermore, diving the accuracy by cost is evident in the benefit of using envirotyping (Figure 4B).
Figure 4. A. Heritability in the optimized MET (OPT_MET) and original scenario (Full_MET). B. Heritability divided by total cost in the OPT_MET and Full_MET scenarios.
Envirotyping’s impact: reshaping breeding strategies
Jumping back to the main question: Is envirotyping beneficial to my breeding program? Yes! Even a simple task like this shows that it is possible to better allocate trials in advance, borrow information from other regions, recommend/identify the best variety for each location, reduce costs, and increase/maintain accuracy. However, the potential is much greater!
Feel free to try the mentioned envirotyping tools by yourself. The EnvRtype and SoilType R libraries can be downloaded at https://github.com/allogamous/EnvRtype and https://github.com/rfn-qtl/SoilType, respectively.
