Using Deep Learning to Identify Grape Resistance to Mildew

Jim Crocker
14th June, 2024

Using Deep Learning to Identify Grape Resistance to Mildew

The OIV 452-1 scoring scale illustrates how increasing grapevine resistance to Plasmopara viticola is characterized by a shift from widespread pathogen sporulation (a, b) to minimal or no sporulation, often with visible necrotic spots as a defense response (c–e).

Image adapted from: Macia et al. / CC BY (Source)

Key Findings

  • Researchers from INRAE, CNRS, and Université de Toulouse developed a machine learning method to identify grapevine resistance to downy mildew
  • The method uses images of infected grapevine leaf discs and a neural network to automate resistance assessment
  • The Swin transformer model achieved 81.7% accuracy in predicting resistance and 97% accuracy in identifying genotype differences, much faster than human experts
Downy mildew, caused by the oomycete Plasmopara viticola, is a significant threat to European grapevines (Vitis vinifera), causing substantial economic losses in viticulture. Traditionally, this disease has been managed through repeated applications of fungicides, but this method is not sustainable due to environmental concerns and the emergence of fungicide-resistant pathogen strains[2][3]. A promising alternative is the development of grapevine varieties with natural resistance to downy mildew. However, the effectiveness of these resistant varieties can be compromised by virulent pathogen strains that overcome the resistance loci[4]. To address this challenge, researchers from INRAE, CNRS, and Université de Toulouse have developed a high-throughput machine learning phenotyping method to identify new resistance loci[1]. The study utilized images of grapevine leaf discs infected with P. viticola, annotated with the OIV 452-1 values. This standard scale, used by experts to visually assess resistance, considers two variables: sporulation (the production of spores) and necrosis (tissue death). By training neural networks with this annotated dataset, the researchers aimed to automate and accelerate the phenotyping process. Various machine learning models were tested, with the Swin transformer encoder achieving the best results. This model demonstrated an accuracy of 81.7% in predicting resistance based on the annotated images. More impressively, it achieved an accuracy of 97% in identifying differences between genotypes, matching human observers but at a throughput rate 650% faster. This innovative approach addresses several limitations of traditional visual assessments, which are time-consuming, low-throughput, and dependent on expert judgment. By automating the phenotyping process, the method developed by the researchers can significantly speed up the identification of new resistance loci, facilitating the breeding of more durable resistant grapevine varieties. The study's findings build on previous research that highlighted the need for sustainable disease control methods in viticulture. For instance, the historical reliance on copper-based fungicides and the emergence of fungicide-resistant strains have underscored the importance of developing alternative strategies[2]. Additionally, the spread of P. viticola from its native North America to Europe and beyond has demonstrated the global nature of this pathogen and the need for coordinated efforts to manage its impact[5]. Moreover, the study's use of machine learning for high-throughput phenotyping aligns with the broader trend of integrating advanced technologies into agricultural research. By leveraging neural networks and annotated datasets, researchers can now process large volumes of data quickly and accurately, identifying resistant genotypes that might have been overlooked using traditional methods. In conclusion, the development of a machine learning phenotyping method by INRAE, CNRS, and Université de Toulouse represents a significant advancement in the fight against grapevine downy mildew. By automating the assessment of disease symptoms, this method can accelerate the breeding of resistant grapevine varieties, providing a more sustainable and effective solution to a longstanding problem in viticulture. The integration of this technology with existing knowledge and breeding programs holds the promise of enhancing the durability of resistance and ensuring the long-term viability of grapevine cultivation.

AgricultureBiotechPlant Science

References

Main Study

1) Phenotyping grapevine resistance to downy mildew: deep learning as a promising tool to assess sporulation and necrosis.

Published 13th June, 2024

https://doi.org/10.1186/s13007-024-01220-4

Related Studies

2) Plasmopara viticola the Causal Agent of Downy Mildew of Grapevine: From Its Taxonomy to Disease Management.

https://doi.org/10.3389/fmicb.2022.889472

3) Variation in Susceptibility to Downy Mildew Infection in Spanish Minority Vine Varieties.

https://doi.org/10.3390/plants12142638

4) Breakdown of resistance to grapevine downy mildew upon limited deployment of a resistant variety.

https://doi.org/10.1186/1471-2229-10-147

5) Europe as a bridgehead in the worldwide invasion history of grapevine downy mildew, Plasmopara viticola.

https://doi.org/10.1016/j.cub.2021.03.009


Related Articles

An unhandled error has occurred. Reload 🗙