Seeing the unseen: New hyperspectral-metabolomics pipeline accelerates salt-tolerant crop screening
image: (A) The pipeline for Multiple Stepwise Regression approach for predicting metabolite concentrations utilizing reflectance spectroscopy with a quintuple-phase methodology: acquisition of original spectra, spectral transformation processing, feature selection, model construction, and parameter inversion. We quantitatively assessed the fractional-order spectral characteristics of M. truncatula leaf hyperspectral data, employing a Sequential Forward Feature Selection algorithm to determine the most effective fractional differential spectral bands for metabolite sensitivity enhancement. Then the optimal band combination was selected to construct a Multiple Stepwise Regression model. A stringent screening was conducted to improve the model's accuracy, identifying content conversion models for 101 metabolites exceeding the threshold (R2 > 0.9), as indicated by the red dotted line in parameter inversion section (Tables S5–S6). (B) Heatmap showing the variations in the content of metabolites with conversion models, obtained through converting hyperspectral data. Details are shown in Table S6. 1 d, 3 d and 5 d stand for salt treatment for 1, 3 and 5 days, respectively. SS and CK represent Jemalong A17 with and without salt treatment. Variation is scaled to range from 0 to 1 and shown by different colors. (C) KEGG pathway enrichment analysis of metabolites with content conversion models. KEGG analysis was performed and shown as described in Fig. 2, Fig. 3F. The color of the columns indicates the P value as shown in Table S7.
Credit: The author
Tested on over 1,000 Medicago truncatula lines, the method tripled the detection rate of salt-tolerant phenotypes and achieved a 90% accuracy, offering a promising tool for accelerating crop breeding in a changing climate.
In modern agriculture, identifying stress-tolerant traits in crops—such as resistance to salt stress—is essential to ensure global food security. Traditional phenotyping methods rely on visible plant traits, which often emerge too late in the stress cycle to be useful for early screening. Meanwhile, hyperspectral imaging provides detailed biochemical information non-invasively, and metabolomics offers insight into internal plant responses. Despite their complementary strengths, these two omics tools have rarely been effectively integrated, limiting their potential in crop breeding.
A study (DOI: 10.1016/j.plaphe.2025.100020) published in Plant Phenomics on 21 February 2025 by Jingyu Zhang’s and Kang Chong’s team, Institute of Botany, Chinese Academy of Sciences, enables early, accurate, and non-destructive identification of salt-tolerant plant phenotypes, offering a powerful tool for precision crop breeding.
To identify salt-tolerant phenotypes with high precision, researchers developed a two-stage screening pipeline that integrates hyperspectral sensing and metabolomic profiling. First, phenomic and metabolomic data were collected from Medicago truncatula plants treated with 200 mM NaCl for 1, 3, and 5 days. Hyperspectral data, captured across 2151 continuous narrow bands (350–2500 nm), were aligned with metabolite profiles from the same plant parts and timepoints. Principal Component Analysis (PCA) of metabolomics revealed strong stress-induced responses, including changes in antioxidant pathways and polyamine biosynthesis. A total of 667 metabolites were associated with salt tolerance, and 122 showed consistent relevance across all timepoints. From these, 51 metabolite-based spectral indices were developed by correlating hyperspectral features with metabolite levels (r > 0.8), allowing metabolic shifts to be detected using spectral data alone. Subsequently, the team constructed metabolite content conversion models for 101 key metabolites using stepwise multiple regression, achieving high accuracy (R² > 0.9). These models enabled the creation of hyperspectral-derived metabolic profiles. The pipeline’s primary screening used random forest models with 84 features—including spectral indices and RGB-derived indices—to classify salt-tolerant candidates, selecting 49 mutants for further analysis. In the secondary screening, converted metabolite profiles were used to reclassify these candidates, ultimately identifying 20 salt-tolerant mutants. These individuals showed no visible differences from the general population but exhibited distinct biochemical signatures, verified by PCA and hierarchical clustering. Follow-up experiments under extended salt stress confirmed 18 of these as true positives, demonstrating a 90% screening accuracy. Comparative analysis of borderline cases validated the pipeline’s capacity to discriminate subtle phenotypic differences, underscoring its potential in early-stage, non-destructive trait selection for precision crop breeding.
This approach enables early phenotype prediction—just five days into salt stress—long before visible symptoms appear. It also reveals underlying biochemical mechanisms and could help breeders target specific metabolic pathways to engineer improved traits.
###
References
DOI
Original Source URL
https://doi.org/10.1016/j.plaphe.2025.100020
Funding information
This work was supported by the Strategic Priority Research Program of Chinese Academy of Sciences (XDA26030102), the CAS-CSIRO Project (063GJHZ2022047MI) and the CAS Special Research Assistant (SRA) Program (Y973RG1001).
About Plant Phenomics
Science Partner Journal Plant Phenomics is an online-only Open Access journal published in affiliation with the State Key Laboratory of Crop Genetics & Germplasm Enhancement, Nanjing Agricultural University (NAU) and distributed by the American Association for the Advancement of Science (AAAS). Like all partners participating in the Science Partner Journal program, Plant Phenomics is editorially independent from the Science family of journals. Editorial decisions and scientific activities pursued by the journal's Editorial Board are made independently, based on scientific merit and adhering to the highest standards for accurate and ethical promotion of science. These decisions and activities are in no way influenced by the financial support of NAU, NAU administration, or any other institutions and sponsors. The Editorial Board is solely responsible for all content published in the journal. To learn more about the Science Partner Journal program, visit the SPJ program homepage.