An illustration of model agnostic explainability methods applied to environmental data.

Historically, two primary criticisms statisticians have of machine learning and deep neural models is their lack of uncertainty quantification and the inability to do inference (i.e., to explain what inputs are important). Explainable AI has developed in the last few years as a sub-discipline of computer science and machine learning to mitigate these concerns (as well as concerns of fairness and transparency in deep modeling). In this article, our focus is on explaining which inputs are important in models for predicting environmental data. In particular, we focus on three general methods for explainability that are model agnostic and thus applicable across a breadth of models without internal explainability: "feature shuffling", "interpretable local surrogates", and "occlusion analysis". We describe particular implementations of each of these and illustrate their use with a variety of models, all applied to the problem of long-lead forecasting monthly soil moisture in the North American corn belt given sea surface temperature anomalies in the Pacific Ocean.

Visual Analytics for the Coronavirus COVID-19 Pandemic.

The coronavirus disease COVID-19 was first reported in Wuhan, China, on December 31, 2019. The disease has since spread throughout the world, affecting 227.2 million individuals and resulting in 4,672,629 deaths as of September 9, 2021, according to the Johns Hopkins University Center for Systems Science and Engineering. Numerous sources track and report information on the disease, including Johns Hopkins itself, with its well-known Novel Coronavirus Dashboard. We were also interested in providing information on the pandemic. However, rather than duplicating existing resources, we focused on integrating sophisticated data analytics and visualization for region-to-region comparison, trend prediction, and testing and vaccination analysis. Our high-level goal is to provide visualizations of predictive analytics that offer policymakers and the general public insight into the current pandemic state and how it may progress into the future. Data are visualized using a web-based jQuery+Tableau dashboard. The dashboard allows both novice viewers and domain experts to gain useful insights into COVID-19's current and predicted future state for different countries and regions of interest throughout the world.

Loci That Control Nonlinear, Interdependent Responses to Combinations of Drought and Nitrogen Limitation.

Crop improvement must accelerate to feed an increasing human population in the face of environmental changes. Including anticipated climatic changes with genetic architecture in breeding programs could better optimize improvement strategies. Combinations of drought and nitrogen limitation already occur world-wide. We therefore analyzed the genetic architecture underlying the response of Zea mays to combinations of water and nitrogen stresses. Recombinant inbreds were subjected to nine combinations of the two stresses using an optimized response surface design, and their growth was measured. Three-dimensional response surfaces were fit globally and to each polymorphic allele to determine which genetic markers were associated with different response surfaces. Three quantitative trait loci that produced nonlinear surfaces were mapped. To better understand the physiology of the response, we developed a model that reproduced the shapes of the surfaces, their most characteristic feature. The model contains two components that each combine the nitrogen and water inputs. The relative weighting of the two components and the inputs is governed by five parameters, and each QTL affects all five parameters.We estimated the model's parameter values for the experimental surfaces using a mesh of points that covered the surfaces' most distinctive regions. Surfaces computed using these values reproduced the experimental surfaces well, as judged by three different criteria at the mesh points. The modeling and shape comparison techniques used here can be extended to other complex, high-dimensional, nonlinear phenotypes. We encourage the application of our findings and methods to experiments that mix crop protection measures, stresses, or both, on elite and landrace germplasm.

Abiotic stress growth conditions induce different responses in kernel iron concentration across genotypically distinct maize inbred varieties.

The improvement of grain nutrient profiles for essential minerals and vitamins through breeding strategies is a target important for agricultural regions where nutrient poor crops like maize contribute a large proportion of the daily caloric intake. Kernel iron concentration in maize exhibits a broad range. However, the magnitude of genotype by environment (GxE) effects on this trait reduces the efficacy and predictability of selection programs, particularly when challenged with abiotic stress such as water and nitrogen limitations. Selection has also been limited by an inverse correlation between kernel iron concentration and the yield component of kernel size in target environments. Using 25 maize inbred lines for which extensive genome sequence data is publicly available, we evaluated the response of kernel iron density and kernel mass to water and nitrogen limitation in a managed field stress experiment using a factorial design. To further understand GxE interactions we used partition analysis to characterize response of kernel iron and weight to abiotic stressors among all genotypes, and observed two patterns: one characterized by higher kernel iron concentrations in control over stress conditions, and another with higher kernel iron concentration under drought and combined stress conditions. Breeding efforts for this nutritional trait could exploit these complementary responses through combinations of favorable allelic variation from these already well-characterized genetic stocks.

Loci controlling nitrate reductase activity in maize: ultraviolet-B signaling in aerial tissues increases nitrate reductase activity in leaf and root when responsive alleles are present.

Environmental factors, such as ultraviolet-B (UV-B) irradiation, have the ability to affect pathways such as nitrogen metabolism. As fixed nitrogen is the keystone mineral nutrient that controls grain crop yield, any alteration in this cycle can be detrimental to plant productivity. Nitrate reductase enzyme activity is responsible for the reduction of nitrate to nitrite, and nitrate is the major form of nitrogen assimilated in plants. In maize (Zea mays L.) production, nitrate assimilation kinetics are important for both high- and low-input agricultural systems. Nitrate reductase protein activity is controlled by phosphatases and kinases. Nitrate reductase activity is responsive to environmental signals such as light-dark cycles and UV-B radiation, although the regulatory controls are not yet fully understood. We have determined the location of maize genetic factors that control nitrate reductase activity and the extent of contribution of each of these factors, both locally in the leaf tissue and via long-distance signaling loci that affect root nitrate reductase activity upon leaf UV irradiation. In the IBM94 recombinant inbred mapping population, the loci controlling regulation of nitrate reductase activity under UV-B map to different positions than the loci controlling nitrate reductase activity in unexposed plants.

Chromosomal loci important for cotyledon opening under UV-B in Arabidopsis thaliana.

BACKGROUND: Understanding of the genetic architecture of plant UV-B responses allows extensive targeted testing of candidate genes or regions, along with combinations of those genes, for placement in metabolic or signal transduction pathways. RESULTS: Composite interval mapping and single-marker analysis methods were used to identify significant loci for cotyledon opening under UV-B in four sets of recombinant inbred lines. In addition, loci important for canalization (stability) of cotyledon opening were detected in two mapping populations. One candidate locus contained the gene HY5. Mutant analysis demonstrated that HY5 was required for UV-B-specific cotyledon opening. CONCLUSIONS: Structured mapping populations provide key information on the degree of complexity in the genetic control of UV-B-induced cotyledon opening in Arabidopsis. The loci identified using quantitative trait analysis methods are useful for follow-up testing of candidate genes.

Maize leaf epiphytic bacteria diversity patterns are genetically correlated with resistance to fungal pathogen infection.

Plant leaves host a specific set of microbial epiphytes. Plant genetic and solar UV-B radiation effects on the diversity of the phyllosphere were examined by measuring epiphytic bacterial ribosomal DNA diversity in a maize recombinant inbred (RI) mapping population. Several chromosomal quantitative trait loci (QTL) with significant effects on bacterial diversity were identified, some of which had effects only in the presence of UV-B radiation and others that had effects both with and without UV-B. Candidate genes with allele-specific effects were mapped to the bacterial diversity chromosomal regions. A glutamate decarboxylase candidate gene was located at a UV-B-specific chromosomal locus, and in a comparison between two RI lines with contrasting bacterial diversity phenotypes, high bacterial diversity was associated with high levels of glutamate decarboxylase enzyme activity, a component of the gamma-aminobutyric acid (GABA) pathway. The bacterial diversity loci exhibited a significant overlap with loci connected with Southern leaf blight (SLB) susceptibility in the field. A SLB-resistant inbred genotype had less beta bacterial diversity, and antibiotic treatment of inbreds increased this diversity. These results suggest that the GABA pathway is genetically associated with phyllosphere bacterial diversity. Furthermore, the colocalization of QTL between low bacterial diversity and fungal blight-resistance and the increase in beta diversity in antibiotic-treated leaves suggest that occupation of leaf habitats by a particular set of suppressive bacteria may restrict phyllosphere bacterial variability and increase resistance to fungal infection.

Coordinated regulation of maize genes during increasing exposure to ultraviolet radiation: identification of ultraviolet-responsive genes, functional processes and associated potential promoter motifs.

Genetic gain in the yield of modern maize reflects increased stress tolerance. The manipulation of genes for deliberate alterations in tolerance relies on an understanding of the regulation and components of stress responses. Transcriptome analysis of an ultraviolet (UV) radiation time course with paired treatment and control measurements yielded groups of coordinately regulated genes and gene ontology processes. A comparison of the patterns of gene expression with patterns of morphological changes allowed the identification of physiologically relevant gene expression regulons. A set of genes significantly affected by UV radiation in maize leaves was selected by linear modelling plus order-restricted inference profile matches. This gene list was used to find upstream sequence motifs that predict the UV regulation of maize gene expression.

Order-restricted inference for ordered gene expression (ORIOGEN) data under heteroscedastic variances.

This article extends the order restricted inference approach for time-course or dose-response gene expression microarray data, introduced by Peddada and colleagues (2003) for the case when gene expression is heteroscedastic over time or dose. The new methodology uses an iterative algorithm to estimate mean expression at various times/doses when mean expression is subject to pre-defined patterns or profiles, known as order-restrictions. Simulation studies reveal that the resulting bootstrap-based methodology for gene selection maintains the false positive rate at the nominal level while competing well with ORIOGEN in terms of power. The proposed methodology is illustrated using a breast cancer cell-line data analyzed by Peddada and colleagues (2003).