Impressions from the 2024 Genotype by Environment by Management (GxExM) Symposium

 

The Genotype by Environment by Management (GxExM) Symposium III took place on October 30-31, 2024, on the Wageningen Campus, NL. Bringing together breeders, agronomists, and farmers, this annual gathering aimed to explore the complexities of crop performance in varied conditions. From climate shifts to cutting-edge machine learning (ML) advancements, this year’s event sparked insightful discussions among experts passionate about improving crop performance and resilience.

Talk by Dr. István Dékány: Interpretable Deep Learning for Multi-Trait Prediction

A highlight of the symposium was István Dékány’s talk, “Accurate Multi-Trait Prediction Across Cycles of Selection by Interpretable Deep Learning”. This talk, the result of a collaboration between Computomics and the University of Queensland under the ARC Centre of Excellence for Plant Success in Nature and Agriculture, introduced innovative applications of machine learning designed specifically for agricultural needs.

István presented the benefits of predictive models that go beyond general-purpose machine learning approaches by integrating biological information on gene-to-phenotype networks. These informed models offer two key advantages. Firstly, they increase forward-prediction accuracy over many selection cycles into the future even under very strong data drift. Secondly, they offer superior interpretability, revealing the genetic causality of intermediate traits and their interactions. This emphasis on a better understanding of selection response is essential, as it ensures that breeders can understand the logic behind predictions and apply them to real-world agricultural scenarios.

Dr. István Dékány
Machine Learning Scientist at Computomics

 

A Collaborative and Knowledge-Driven Community

The atmosphere at this year’s symposium was collaborative, with participants eager to exchange ideas and to network. Attendees shared a strong sense of community, recognizing the importance of collective insights to tackle the challenges facing agriculture today. Many of them, well-acquainted through previous conferences and ongoing projects, were open to discussing their work and exploring cross-disciplinary approaches to problem-solving.

István’s talk resonated particularly well with the audience, sparking numerous discussions around the potential of AI to transform plant breeding and crop management. Attendees expressed interest in informed models and other innovative approaches to AI, noting how AI can drive precision agriculture forward by integrating science-backed models with data-driven methods.

Advanced AI Development: Shaping The Future of Crop Performance Prediction

Two major directions of active research dominated the talks and ensuing discussions about AI: transfer learning and informed models. Transfer learning involves pre-training predictive models on large datasets, then fine-tuning them for related new learning tasks, which results in superior predictive ability with respect to traditional ML in situations where data are scarce. In contrast, informed models are designed from the ground up by incorporating domain knowledge, e.g.Im biological networks, crop growth patterns, or biophysical properties into their architectures, allowing them to make more accurate predictions.

Addressing the Challenges of Climate Change

Climate change was a recurring topic at the symposium, reflecting the urgent need for agriculture to adapt to increasingly volatile weather patterns. Beyond discussing considering climate change as increased heat stress on plants, conversations shifted toward the need of breeding plants that are resilient against the increasing danger of pests and viruses, as well as erratic weather. Practical solutions, such as intercropping, that can help maintain biodiversity and mitigate the sector’s carbon footprint, were also discussed. Attendees also shared concerns about potential extreme climate scenarios, including the consequences of a weakening Atlantic Meridional Overturning Circulation (AMOC), which would lead to cooling in Europe.

This emphasis on resilience underlines why GxExM approaches are so critical in today’s agricultural landscape. Understanding and predicting crop responses under varying climate conditions allows breeders and agronomists to proactively develop strategies that could sustain crop performance amid growing climate instability.

Building Connections with Industry: New Collaborations and Networking

Throughout the event, István Dékány engaged with multiple industry representatives interested in exploring how AI can benefit their specific contexts. This symposium offered a valuable platform for academic and industry players to connect, fostering an environment where new collaborations could emerge to tackle real-world challenges in agriculture.

Closing Thoughts

The 2024 GxExM Symposium III successfully brought together a community of experts committed to advancing crop performance through innovative science and technology. With a focus on AI and climate resilience, the discussions at this year’s symposium underscored the importance of predictive models that can adapt to changing conditions and complex interactions.

If you are interested in hearing more about István’s talk and our collaboration with the University of Queensland under the ARC Centre of Excellence for Plant Success in Nature and Agriculture, reach out to István directly at istvan.dekany@computomics.com.

 

Image: Field in Burgundy, France - by Lucas Marconnet on Unsplash