Leaf hyperspectral phenotyping is widely used for non-destructive trait estimation in plants, but datasets with limited biological samples relative to high-dimensional spectral variables remain methodologically challenging because of strong band redundancy, limited robustness under developmental shift, and weak interpretability of feature-importance analyses. Here we present a reproducible and interpretable workflow for SPAD-based leaf chlorophyll status phenotyping from hyperspectral reflectance, using oat (Avena sativa L.) as a representative use case. The workflow integrates spectral preprocessing benchmarking, two-stage band selection (Lasso screening followed by successive projections refinement), tabular deep learning optimized for structured spectral predictors (TabM-v2 with Mixup and stochastic weight averaging), and a multi-level interpretation scheme spanning feature, region, and dependency analyses. Using 1,142 retained ROI-level modeling samples derived from 200 sampled oat leaves (125 bands, 400–1000 nm) and five independent growth-stage validation batches, we show that the workflow achieves strong predictive performance ( R^2 = 0.8575 ± 0.0021 ) while reducing spectral dimensionality by 80
Improving root system architecture (RSA) requires rapid and practical field-based phenotyping methods. We established a simple vertical root-pulling force (RPF) protocol for characterizing wheat cybrids under the Tottori sandy field conditions. RPF ranged from 1.47–20.61 kg in mature plants and 0.58–6.79 kg in immature plants. The optimized protocol enabled measurements at approximately 5 s per plant and over 1000 plants within 24 working hours for four concurrent operators, while allowing recovery of most of the visible root system with little visible breakage. The cybrids showed substantial variation in RSA-related traits. RPF was most reliable at maturity, with repeatability of 0.83 and significant associations with root area, root biomass, and root-adhering soil mass. In immature plants, RPF repeatability was lower (0.13) and the line effect was not significant (p = 0.052). The optimized RPF protocol provides a rapid initial screening approach for mature wheat under the uniform sandy field conditions evaluated. RPF does not directly measure complete root system architecture, and root breakage or incomplete recovery cannot be excluded. Projected 2D root area and other measurements of extracted roots should therefore be interpreted as characteristics of the recovered root fraction. Selected lines can subsequently be examined using more detailed root-phenotyping methods when additional architectural information is required.
Extreme weather events, increasing water scarcity, and rising temperatures pose major challenges to agricultural sustainability and global food security. Developing crop varieties with enhanced stress resilience is therefore a priority, yet conventional breeding remains labor-intensive, time-consuming, and costly. Integrating innovative sensing technologies into breeding pipelines offers a promising strategy to improve the efficiency and precision of genotype selection. Here, we present two case studies demonstrating the potential of the Bioristor, an organic electrochemical transistor (OECT)-based in vivo sensor, to support phenotypic selection under drought stress. In the first study, three commercial tomato (Solanum lycopersicum L.) varieties (Mariner, Faber, and Red Setter) were continuously monitored during drought and recovery. In the second study, the Bioristor was applied to a panel of tomato introgression lines (ILs), enabling continuous, real-time characterization of physiological responses to drought and combined drought–heat stress. Bioristor-derived measurements enabled the robust classification of both commercial varieties and ILs according to their stress tolerance. Novel sensor-derived indices enhanced genotype discrimination by revealing differences in the temporal dynamics of physiological responses, particularly those associated with water and ion transport. Beyond improving phenotypic selection, continuous in vivo monitoring provided new insights into stress onset, acclimation, and recovery, identifying genotype-specific response patterns that are not captured by conventional phenotyping approaches. These findings demonstrate the potential of the Bioristor as a high-resolution functional phenotyping tool for early genotype classification and breeding. By complementing genomic and automated phenotyping platforms with continuous, in vivo physiological information, the Bioristor provides a deeper understanding of plant stress responses and represents a promising technology for developing more resilient crop varieties.
Grain-level three-dimensional phenotyping of rice provides quantitative traits for breeding and yield assessment, but remains challenging because dense panicles exhibit severe organ adhesion, boundary ambiguity, and large-scale point-cloud complexity. We developed PRSNeRF, a grain-level phenotyping workflow that combines contour-guided neural radiance field reconstruction with point cloud instance segmentation, and evaluated it on rice grown in controlled indoor environments. PRSNeRF integrates CoNeRF for reconstruction and PRSNet for segmentation. CoNeRF combines image deblurring, contour-guided sampling, and rendering-projection consistency to improve geometric fidelity and suppress floating artifacts. PRSNet uses a Mamba-based encoder, adaptive Gaussian detail enhancement, multi-scale feature fusion, and three-dimensional dynamic anchor boxes to separate densely adjacent grains. Under a fixed 5,000-iteration budget on the Blur Rice benchmark, CoNeRF reduced the scaled NeRF-space Chamfer distance from 10.61 to 7.81. Relative to the 30,000-iteration NeRFacto baseline, it shortened training time from 368 s to 77 s and point cloud export time from 120 s to 10 s. On the RiceNeRF-IS dataset, plant-level five-fold cross-validation showed that PRSNet achieved an overall AP of 87.31
Batch processing of multiple genome assemblies with EDTA produces numerous result directories that require inspection before downstream analysis. Checking expected output files, reported TE content, and TE-library file sizes individually can be time-consuming. A utility that consolidates these checks can help users identify outputs requiring further review. Here, we present EDTA-Audit, a deterministic post-processing screening framework for EDTA outputs. We evaluated EDTA-Audit v1.0.0 using a dataset of 142 plant genome assemblies processed in a concurrent batch-processing setting. Within this dataset, a subset of 13 assemblies was used as a pilot case set to examine the behavior of the default Plant Mode thresholds, and audit outcomes were then summarized across the full dataset. EDTA-Audit assigned PASS or FAIL outcomes and recorded diagnostic reasons for flagged outputs, helping users identify missing required files and unusual TE-content or TE-library profiles requiring troubleshooting or manual inspection. EDTA-Audit consolidates file checks and taxon-informed profile warnings into batch reports, providing a practical first-pass screen to prioritize EDTA outputs for troubleshooting and manual review. EDTA-Audit v1.0.0, the version used for the analyses reported in this study, was implemented in Bash. The repository has since been updated to a Python-based v2.0.0 implementation, developed with the aim of improving usability, maintainability, and extensibility. All results reported here were obtained using v1.0.0.
