Search

X-ray micro-computed tomography (X-ray μCT) has enabled the characterization of the properties and processes that take place in plants and soils at the micron scale. Despite the widespread use of this advanced technique, major limitations in both hardware and software limit the speed and accuracy of image processing and data analysis. Recent advances in machine learning, specifically the application of convolutional neural networks to image analysis, have enabled rapid and accurate segmentation of image data. Yet, challenges remain in applying convolutional neural networks to the analysis of environmentally and agriculturally relevant images. Specifically, there is a disconnect between the computer scientists and engineers, who build these AI/ML tools, and the potential end users in agricultural research, who may be unsure of how to apply these tools in their work. Additionally, the computing resources required for training and applying deep learning models are unique, more common to computer gaming systems or graphics design work, than to traditional computational systems. To navigate these challenges, we developed a modular workflow for applying convolutional neural networks to X-ray μCT images, using low-cost resources in Google's Colaboratory web application. Here we present the results of the workflow, illustrating how parameters can be optimized to achieve best results using example scans from walnut leaves, almond flower buds, and a soil aggregate. We expect that this framework will accelerate the adoption and use of emerging deep learning techniques within the plant and soil sciences.

Basic quantitative and population genetics topics are typically taught in introductory plant breeding courses and are critical for success in upper-level study. Active learning, including simulations and games, may be useful for instruction of these concepts, which rely heavily on theory and may be more challenging for students. The statistical computing language "R" is now routinely used in the analysis of plant breeding experiments, but the command-line interface of the language may be unsuitable for an introductory course. Here we describe "qgshiny" (quantitative genetics in "shiny"), an interactive application for performing simulations to understand basic theory in quantitative and population genetics. The initial version of the application includes modules on three core topics in quantitative genetics: randomly mating populations, genetic variance, and response to selection. Students can specify parameters and initiate simulations to assess their impact on responses such as allele frequency, genetic variance, and genetic gain, which together can be used to reinforce more general learning objectives. Feedback collected from students after engaging with the application suggests this tool can have a positive impact on student learning. The application is bundled in an R package, "qgshiny," which is available through the Comprehensive R Archive Network (CRAN), on GitHub (https://github.com/neyhartj/qgshiny), or interactively through the shinyapps.io platform (http://neyhartj.shinyapps.io/qgshiny).

Crop populations have enormous impacts on agricultural productivity, yet decelerating gains from breeding suggest that selection strategies need to be reconsidered to better align priorities of breeders and growers. Breeders benefit from releasing broadly adapted varieties that perform acceptably well across their target region; growers benefit from selecting a variety that specializes in their specific location. We tested whether these interests are compatible using 182 entries in a mega‐population of malting barley (spring, two‐row, multi‐environment trial [S2MET] population; Hordeum vulgare L.), which was grown across the northern United States. We assessed the strength of genetic–environment interactions (G×E), quantified local adaptation benefits, and tested whether local adaptation and genetic yield potential were correlated. Breeding programs favored broad adaptation (p < 0.001). Still, 63% of entries (114) were among the top 10%, best performing at some locations, and among the worst 10% at others. Some of the best overall entries were specialists, performing especially well in their home locations, and in general, genetic potential and local specialization were positively correlated (p < 0.001). These results suggest that breeding for local performance and broad performance are mutually supporting goals. Releasing broadly excellent, locally exceptional varieties may accelerate genetic gain to meet projected global agricultural demand. Core Ideas: Releasing locally adapted varieties can accelerate genetic gain at the cost of breeding efficiency.We test whether locally and broadly adapted strategies are synergistic in the modern barley mega‐population, spring, two‐row, multi‐environment trial malting barley population (S2MET).Specialist lines tended to yield higher overall, indicating a synergy between these strategies.Broad adaptation at the population level indicated that breeders favored generalist lines.Releasing broadly excellent, locally exceptional varieties is feasible and better aligns breeder and grower needs. [ABSTRACT FROM AUTHOR]

Plant breeding programs expend significant resources on multilocation testing to evaluate genotypes for advancement or potential cultivar release. The selection of genotype entries for these trials is typically based on previous phenotypic data or predictions; yet locations, important contributors to nongenetic variation, are often chosen in a less data‐driven manner. Using agronomic and quality trait data from two long‐term regional barley (Hordeum vulgare L.) nurseries, our objectives were (a) to measure the precision, repeatability, and representativeness of test locations based on multitrait data and (b) to optimize the selection of test locations for use in future trials. When considering traits individually, ideal locations could be identified simply, but a combined analysis of 11 traits indicated that very few locations were broadly favorable, and considerable tradeoffs are necessary. We developed a flexible optimization procedure to select the locations based on their precision, repeatability, and representativeness for multiple traits while simultaneously constraining the total number of locations. Optimization led to a 58–75% reduction in the number of locations, and therefore phenotyping costs, with little loss in data utility. Importantly, our approach allowed locations to be selected for phenotyping different sets of traits (e.g., either agronomic or both agronomic and malting quality), mimicking the often nested structure of trait data collection. This approach may be useful for individual plant breeding programs or collaborations wishing to increase the resource efficiency of these important regional evaluation trials. Core Ideas: Test locations in regional barley breeding nurseries were evaluated for multiple traits.No single location was ever superior for genotypic evaluation of many traits.Optimization identified a reduced set of locations with little loss in phenotypic data utility. [ABSTRACT FROM AUTHOR]

In plant breeding, the goal of genomewide selection is to predict the merit of unobserved individuals, particularly those in the next breeding generation. Predictions of these individuals in unobserved or future environments would be of additional use to a breeder. For many of the complex traits targeted in breeding, this may require management of genotype × environment interactions by, for example, using data from homogeneous groups of environments. Our objectives were to assess the accuracy of genomewide predictions in unobserved environments both within and between breeding generations; we aimed to compare training sets that included data from all possible environments with those that included data from (a) decreasingly similar environments or (b) discrete clusters of similar environments. A 183‐line spring barley (Hordeum vulgare L.) training population and 50‐line offspring test population were phenotyped in 29 location–year environments for grain yield, heading date, and plant height. Environmental similarities were measured using phenotypic data, geographic distance, or environmental covariables. When using training data from more, but decreasingly similar environments, prediction accuracy increased, but marginal gains declined; in some cases, accuracy declined with additional data. Clusters of environments informed by phenotypes (i.e., phenotypic correlations or multiplicative models) typically improved prediction accuracy within a generation, but not between generations (offspring population). Our study suggests that, as an alternative to using data from all available environments, informative subsets may be advantageous for genomewide predictions within a single breeding generation, but not between generations. [ABSTRACT FROM AUTHOR]