By combining these refined phenotypic data with genome-wide association studies (GWAS) and transcriptome sequencing, researchers identified GhDR_UAV1, a previously unknown gene that positively regulates cotton leaf abscission.
Efficient pre-harvest defoliation is essential for modern cotton production, particularly in regions such as China’s Xinjiang, where large-scale mechanized harvesting dominates and relies on clean, uniform leaf drop to maintain efficiency and fiber quality. Although chemical defoliants are widely applied, their environmental, economic, and health risks underscore the importance of breeding varieties capable of strong natural or chemically induced leaf abscission. Yet progress in identifying defoliation-related genes has been constrained by traditional manual field surveys, which are slow, labor-intensive, and error-prone, yielding low-throughput phenotypic data that are insufficient for genome-wide association studies. These limitations have created an urgent need for scalable, precise, and synchronized phenotyping strategies to advance gene discovery and accelerate the development of defoliation-prone cotton cultivars.
A study (DOI: 10.1016/j.plaphe.2025.100109) published in Plant Phenomics on 17 September 2025 by Zuoren Yang’s team, Chinese Academy of Agricultural Sciences, provides a powerful new foundation for breeding defoliation-sensitive cotton lines.
In this study, cotton defoliation was assessed through an integrated high-throughput framework that combined top-down UAV multispectral scanning with bottom-up leaf area index (LAI) measurements, enabling rapid and precise field phenotyping across large populations. This dual-sensor strategy was evaluated against traditional manual surveys to quantify improvements in efficiency and accuracy. To identify the most informative predictors of defoliation, hierarchical segmentation analysis was applied to nine spectral indices and LAI across three statistical levels: leaf number (LN), leaf number difference (LND), and defoliation rate (DR). These selected indices were then used to train four deep learning architectures—1D-CNN, BiGRU, CNN-BiGRU, and CNN-BiGRU-Attention—for trait inversion. Model interpretability was further examined using SHAP analysis. Finally, model-derived phenotypes were incorporated into GWAS to identify genetic loci associated with defoliation, followed by transcriptome analysis and functional validation of candidate genes. Corresponding results demonstrated dramatic gains in efficiency: manual assessment of 214 plots would require 33 days for one investigator, whereas UAV flight required only 20 minutes and LAI measurements four days. UAV and LAI surveys also maintained high accuracy and, when combined, outperformed manual surveys for GWAS applicability. Hierarchical segmentation consistently identified MTCI, VDVI, CI, and LAI as core predictors. Among tested models, the CNN-BiGRU-Attention architecture achieved the highest accuracy (R² > 0.85), particularly for LN and LND. SHAP analysis revealed VDVI as the most influential feature, with chlorophyll-related indices (MTCI, CI) also contributing strongly. GWAS using LND-based predictions showed strongest concordance with manual measurements, identifying significant loci on chromosomes A08 and A10. Transcriptome profiling and functional assays confirmed GhDR_UAV1 as a key TDZ-responsive gene whose overexpression accelerates leaf wilting and abscission, establishing it as a positive regulator of cotton defoliation.
This new phenomics-genomics framework offers cotton breeders a scalable alternative to manual defoliation surveys, reducing weeks of labor to minutes of UAV flight time and enabling genetic studies that were previously infeasible. The discovery of GhDR_UAV1 provides a valuable molecular target for developing cotton varieties that defoliate rapidly and uniformly, reducing reliance on chemical defoliants and lowering production costs while improving fiber purity during mechanical harvesting. Beyond cotton, this approach demonstrates how AI-enabled remote sensing can transform the study of complex field traits, accelerating genomics-assisted breeding in other major crops.
###
References
DOI
10.1016/j.plaphe.2025.100109
Original Source URl
https://doi.org/10.1016/j.plaphe.2025.100109
Funding information
This work was supported by National Key R&D Program of China (2022YFF1001400), National Natural Science Foundation of China (32360509), Natural Science Foundation of Xinjiang Uygur Autonomous Region (2024D01A150), Corps of agricultural science and technology innovation project special (NCG202316), Key Research and Development Program of Xinjiang (2022B02052).
About Plant Phenomics
Plant Phenomics is dedicated to publishing novel research that will advance all aspects of plant phenotyping from the cell to the plant population levels using innovative combinations of sensor systems and data analytics. Plant Phenomics aims also to connect phenomics to other science domains, such as genomics, genetics, physiology, molecular biology, bioinformatics, statistics, mathematics, and computer sciences. Plant Phenomics should thus contribute to advance plant sciences and agriculture/forestry/horticulture by addressing key scientific challenges in the area of plant phenomics.