Innovative use of hyperspectral data and DCGAN newswise

Newswise – A research team used hyperspectral data and Deep Convolution Generative Adversarial Networks (DCGAN) to improve the accuracy of rice grain protein content (GPC) estimation. By generating simulated data, they enhanced the model’s performance, achieving a R² of 0.58 and RRMSE of 6.70%. This technique identified genetic loci including GPC osmtssb1l Jean. This study demonstrates the potential of combining hyperspectral technology and DCGAN for efficient genetic analysis and selection of high-quality rice varieties, paving the way for improved agricultural practices.

Rice (oryza sativa L.) is an important crop feeding more than half of the global population. The demand for high-quality, protein-rich rice is increasing, making accurate grain protein content (GPC) estimation important for breeding improved varieties. Despite advances in genomic tools such as GWAS, traditional phenotyping remains labor-intensive and expensive, posing a barrier. Recent developments in optical and spectral imaging provide high-throughput phenotyping solutions. However, small and unbalanced datasets limit model performance and generalization.

A Study (DOI: 10.34133/plantfenomics.0200) published in plant phenomics 29 May 2024, aims to address these issues by using DCGAN to generate simulated data, increase GPC model accuracy, and explore gene dissection potential.

The research employed hyperspectral data and DCGAN to improve estimation of rice GPC. Raw and normalized spectral data revealed specific absorption features important for GPC analysis. The simulated data generated by DCGAN after 8,000 epochs closely matches the measured data, increasing model accuracy. Partial least squares regression (PLSR) models using these features achieved high validation accuracy (R² = 0.58, RRMSE = 6.70%). Additionally, genome-wide association study (GWAS) analysis with simulated data identified significant SNPs, including osmtssb1l Gene associated with grain storage protein. This approach demonstrates the potential of high-generalization GPC models, facilitating advanced genetic analysis and breeding of rice varieties.

According to Hengbiao Zheng, the study’s lead researcher, “This study provides a new technique for efficient genetic study of phenotypic traits in rice based on hyperspectral technology.”

In summary, the study developed a method using DCGAN to enhance the estimation of rice GPC through hyperspectral data. This approach demonstrates the potential of integrating DCGAN and hyperspectral technology to improve crop phenotyping and genetic analysis. Looking ahead, further refinement and validation in diverse ecological sites and more comprehensive datasets will increase the robustness and applicability of this method, leading to more accurate and efficient breeding of high-quality rice varieties.

,

Reference

DOI

10.34133/plantphenomics.0200

original source url

https://doi.org/10.34133/plantfenomics.0200

Author

hengbiao zheng^1,2*weige tang^2,3*tao yang^1*Meng Zhou¹Kelly Guo¹Tao Cheng¹do waxing¹yan zhu¹yunhui zhang^2,3xia yao¹

Affiliation

1 National Engineering and Technology Center for Information Agriculture (NETCIA), MARA Key Laboratory of Crop System Analysis and Decision Making, MOE Engineering Research Center of Smart Agriculture, Jiangsu Key Laboratory for Information Agriculture, Institute of Smart Agriculture, Nanjing Agricultural University, Nanjing , Jiangsu, China

2.Zhongshan Biological Breeding Laboratory, ZSBBL

3 Provincial Key Laboratory of Agricultural Biology, Institute of Germplasm Resources and Biotechnology, Jiangsu Academy of Agricultural Sciences, Nanjing, Jiangsu, China

*These authors contributed equally.

Funding Information

This work was supported by the National Key Research and Development Program of China (2022YFD2001100), National Natural Science Foundation of China (32101617), Fundamental Research Funds for the Central Universities (JSJL2023005), Zhongshan Biological Reproduction Laboratory (ZSBBL-KY2023) . -05), the key independent research project of the Jiangsu Key Laboratory of Information Agriculture (KLIAZZ2301), and the Jiangsu Collaborative Innovation Center for Modern Crop Production (JCICMCP). We would also like to thank the anonymous reviewers who provided useful comments to improve the manuscript.

About this plant phenomics

plant phenomics is an open access journal published in collaboration with the State Key Laboratory of Crop Genetics and Germplasm Cultivation, Nanjing Agricultural University (NAU) and published by the American Association for the Advancement of Science (AAAS). Like all partners participating in the Science Partner Journal program, plant phenomics Editorially it is independent from the Science family of magazines. Editorial decisions and scientific activities adopted by the Journal’s Editorial Board are carried out independently, on the basis of scientific merit and adhering to the highest standards for the accurate and ethical promotion of science. These decisions and activities are in no way influenced by the financial support of NAU, the NAU administration, or any other institutions and sponsors. The Editorial Board is solely responsible for all content published in the magazine. To learn more about the Science Partner Journal Program, visit the SPJ program homepage.