An official website of the United States government
Abstract Estimates of plant traits derived from hyperspectral reflectance data have the potential to efficiently substitute for traits, which are time or labor intensive to manually score. Typical workflows for estimating plant traits from hyperspectral reflectance data employ supervised classification models that can require substantial ground truth datasets for training. We explore the potential of an unsupervised approach, autoencoders, to extract meaningful traits from plant hyperspectral reflectance data using measurements of the reflectance of 2151 individual wavelengths of light from the leaves of maize (Zea mays) plants harvested from 1658 field plots in a replicated field trial. A subset of autoencoder‐derived variables exhibited significant repeatability, indicating that a substantial proportion of the total variance in these variables was explained by difference between maize genotypes, while other autoencoder variables appear to capture variation resulting from changes in leaf reflectance between different batches of data collection. Several of the repeatable latent variables were significantly correlated with other traits scored from the same maize field experiment, including one autoencoder‐derived latent variable (LV8) that predicted plant chlorophyll content modestly better than a supervised model trained on the same data. In at least one case, genome‐wide association study hits for variation in autoencoder‐derived variables were proximal to genes with known or plausible links to leaf phenotypes expected to alter hyperspectral reflectance. In aggregate, these results suggest that an unsupervised, autoencoder‐based approach can identify meaningful and genetically controlled variation in high‐dimensional, high‐throughput phenotyping data and link identified variables back to known plant traits of interest.
more »
« less
Semantic Segmentation of Sorghum Using Hyperspectral Data Identifies Genetic Associations
This study describes the evaluation of a range of approaches to semantic segmentation of hyperspectral images of sorghum plants, classifying each pixel as either nonplant or belonging to one of the three organ types (leaf, stalk, panicle). While many current methods for segmentation focus on separating plant pixels from background, organ-specific segmentation makes it feasible to measure a wider range of plant properties. Manually scored training data for a set of hyperspectral images collected from a sorghum association population was used to train and evaluate a set of supervised classification models. Many algorithms show acceptable accuracy for this classification task. Algorithms trained on sorghum data are able to accurately classify maize leaves and stalks, but fail to accurately classify maize reproductive organs which are not directly equivalent to sorghum panicles. Trait measurements extracted from semantic segmentation of sorghum organs can be used to identify both genes known to be controlling variation in a previously measured phenotypes (e.g., panicle size and plant height) as well as identify signals for genes controlling traits not previously quantified in this population (e.g., stalk/leaf ratio). Organ level semantic segmentation provides opportunities to identify genes controlling variation in a wide range of morphological phenotypes in sorghum, maize, and other related grain crops.
more »
« less
- Award ID(s):
- 1557417
- PAR ID:
- 10338659
- Date Published:
- Journal Name:
- Plant Phenomics
- Volume:
- 2020
- ISSN:
- 2643-6515
- Page Range / eLocation ID:
- 1 to 11
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Context: Stalk lodging causes up to 43 % of yield losses in maize (Zea mays L.) worldwide, significantly worsening food and feed shortages. Stalk lodging resistance is a complex trait specified by several structural, material, and geometric phenotypes. However, the identity, relative contribution, and genetic tractability of these intermediate phenotypes remain unknown. Objective: The study is designed to identify and evaluate plant-, organ-, and tissue-level intermediate phenotypes associated with stalk lodging resistance following standardized phenotyping protocols and to understand the variation and genetic tractability of these intermediate phenotypes. Methods: We examined 16 diverse maize hybrids in two environments to identify and evaluate intermediate phenotypes associated with stalk flexural stiffness, a reliable indicator of stalk lodging resistance, at physiological maturity. Engineering-informed and machine learning models were employed to understand relationships among intermediate phenotypes and stalk flexural stiffness. Results: Stalk flexural stiffness showed significant genetic variation and high heritability (0.64) in the evaluated hybrids. Significant genetic variation and comparable heritability for the cross-sectional moment of inertia and Young’s modulus indicated that geometric and material properties are under tight genetic control and play a combinatorial role in determining stalk lodging resistance. Among the twelve internode-level traits measured on the bottom and the ear internode, most traits exhibited significant genetic variation among hybrids, moderate to high heritability, and considerable effect of genotype × environment interaction. The marginal statistical model based on structural engineering beam theory revealed that 74–80 % of the phenotypic variation for flexural stiffness was explained by accounting for the major diameter, minor diameter, and rind thickness of the stalks. The machine learning model explained a relatively modest proportion (58–62 %) of the variation for flexural stiffness.more » « less
-
Abstract Most semantic segmentation approaches of big data hyperspectral images use and require preprocessing steps in the form of patching to accurately classify diversified land cover in remotely sensed images. These approaches use patching to incorporate the rich spatial neighborhood information in images and exploit the simplicity and segmentability of the most common datasets. In contrast, most landmasses in the world consist of overlapping and diffused classes, making neighborhood information weaker than what is seen in common datasets. To combat this common issue and generalize the segmentation models to more complex and diverse hyperspectral datasets, in this work, we propose a novel flagship model: Clustering Ensemble U-Net. Our model uses the ensemble method to combine spectral information extracted from convolutional neural network training on a cluster of landscape pixels. Our model outperforms existing state-of-the-art hyperspectral semantic segmentation methods and gets competitive performance with and without patching when compared to baseline models. We highlight our model’s high performance across six popular hyperspectral datasets including Kennedy Space Center, Houston, and Indian Pines, then compare them to current top-performing models.more » « less
-
Abstract An early event in plant organogenesis is establishment of a boundary between the stem cell containing meristem and differentiating lateral organ. In maize (Zea mays), evidence suggests a common gene network functions at boundaries of distinct organs and contributes to pleiotropy between leaf angle and tassel branch number, two agronomic traits. To uncover regulatory variation at the nexus of these two traits, we use regulatory network topologies derived from specific developmental contexts to guide multivariate genome-wide association analyses. In addition to defining network plasticity around core pleiotropic loci, we identify new transcription factors that contribute to phenotypic variation in canopy architecture, and structural variation that contributes tocis-regulatory control of pleiotropy between tassel branching and leaf angle across maize diversity. Results demonstrate the power of informing statistical genetics with context-specific developmental networks to pinpoint pleiotropic loci and theircis-regulatory components, which can be used to fine-tune plant architecture for crop improvement.more » « less
-
Abstract Plant architecture is a major determinant of planting density, which enhances productivity potential for crops per unit area. Genomic prediction is well positioned to expedite genetic gain of plant architectural traits since they are typically highly heritable. Additionally, the adaptation of genomic prediction models to query predictive abilities of markers tagging certain genomic regions could shed light on the genetic architecture of these traits. Here, we leveraged transcriptional networks from a prior study that contextually described developmental progression during tassel and leaf organogenesis in maize (Zea mays) to inform genomic prediction models for architectural traits. Since these developmental processes underlie tassel branching and leaf angle, 2 important agronomic architectural traits, we tested whether genes prioritized from these networks quantitatively contribute to the genetic architecture of these traits. We used genomic prediction models to evaluate the ability of markers in the vicinity of prioritized network genes to predict breeding values of tassel branching and leaf angle traits for 2 diversity panels in maize and diversity panels from sorghum (Sorghum bicolor) and rice (Oryza sativa). Predictive abilities of markers near these prioritized network genes were similar to those using whole-genome marker sets. Notably, markers near highly connected transcription factors from core network motifs in maize yielded predictive abilities that were significantly greater than expected by chance in not only maize but also closely related sorghum. We expect that these highly connected regulators are key drivers of architectural variation that are conserved across closely related cereal crop species.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.34133/2020/4216373
