Integrating genomics, phenomics, and deep learning improves the predictive ability for Fusarium head blight-related traits in winter wheat

Authors: THAPA, SUBASH - North Dakota State University; GILL, HARSIMARDEEP - North Dakota State University; HALDER, JYOTIRMOY - North Dakota State University; RANA, ANSHUL - North Dakota State University; ALI, SHAUKAT - North Dakota State University; MAIMAITIJIANG, MAITINIYAZI - North Dakota State University; GILL, UPINDER - North Dakota State University; Bernardo, Amy; St Amand, Paul; Bai, Guihua; SEHGAL, SUNISH - North Dakota State University

Submitted to: The Plant Genome
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 4/14/2024
Publication Date: 6/9/2024
Citation: Thapa, S., Gill, H., Halder, J., Rana, A., Ali, S., Maimaitijiang, M., Gill, U., Bernardo, A.E., St Amand, P.C., Bai, G., Sehgal, S. 2024. Integrating genomics, phenomics, and deep learning improves the predictive ability for Fusarium head blight-related traits in winter wheat. The Plant Genome. Article.e20470. https://doi.org/10.1002/tpg2.20470.
DOI: https://doi.org/10.1002/tpg2.20470

Interpretive Summary: Fusarium head blight (FHB) is a destructive wheat disease. Visual estimation of FHB resistance traits such as Fusarium damaged kernel (FDK) and deoxynivalenol (DON) is prone to human-biases, which hinders the progress in breeding for FHB resistant cultivars. We used an artificial intelligence (AI) -based method to estimate FDK and used the AI-based FDK in genomic selection (GS). AI-based FDK significantly improved heritability and showed a two-fold increase in prediction accuracy (PA) compared to GS using traditional FDK. The AI-based FDK was used to predict (DON). Inclusion of AI-based FDK with heading days as covariates improved PA for DON by 58% over original method. The PA for DON using selected bands derived from hyperspectral wavelength in multi-trait (MT) models surpassed the single-trait GS model by around 40%. Moreover, we evaluated phenomic selection by integrating hyperspectral imaging with deep-learning (DL) models to directly estimate DON in FHB-infected wheat kernels and observed a PA (R2 = 0.45) comparable to best-performing MT GS models (0.49). This study demonstrates the potential application of AI and vision-based platforms to improve PA for FHB-related traits.

Technical Abstract: Fusarium head blight (FHB) remains one of the most destructive diseases of wheat, causing considerable losses in yield and end-use quality. Phenotyping of FHB resistance traits, Fusarium-damaged kernels (FDK) and deoxynivalenol (DON), are either prone to human-biases or resource expensive, hindering the progress in breeding for FHB resistant cultivars. Though genomic selection (GS) can be an effective way to select for these traits, inaccurate phenotyping remains a hurdle in exploiting this approach. Here, we used an artificial intelligence (AI)-based precise FDK estimation that exhibits high heritability and correlation with DON. Further, GS using AI-based FDK (FDK_QVIS/FDK_QNIR) showed a two-fold increase in prediction accuracy (PA) compared to GS for traditionally estimated FDK (FDK_V). Next, the AI-based FDK was evaluated along with other traits in multi-trait (MT) GS models for DON prediction. Inclusion of FDK_QNIR and FDK_QVIS with heading days as covariates improved PA for DON by 58% over baseline single-trait GS model. We next used hyperspectral imaging of FHB-infected wheat kernels as a novel avenue to improve the MT GS for DON. The PA for DON using selected bands derived from hyperspectral wavelength in MT models surpassed the single-trait GS model by around 40%. Moreover, we evaluated phenomic selection by integrating hyperspectral imaging with deep-learning (DL) models to directly estimate DON in FHB-infected wheat kernels and observed a PA (R2 = 0.45) comparable to best-performing MT GS models (0.49). This study demonstrates the potential application of AI and vision-based platforms to improve PA for FHB-related traits using genomic and phenomic selection.