We are seeing the emergence of a new breeding system that can predict the likely paddock performance of breeding material based on a marriage of biological data, climate records and machine-based artificial intelligence (AI). The new system is called prediction-based crop improvement.
Nuances in how this system works mean that Australia generally - and the Queensland Alliance for Agriculture and Food Innovation (QAAFI) in particular - are regarded as having a 'cutting-edge' advantage in exploiting this development and accelerating genetic gain that benefits grain production.
At the centre of the development work is the University of Queensland's Professor Graeme Hammer, who explains that the key to integrating diverse data sets - genetic, phenotypic, agronomic and climatic - is computer models that can simulate crop growth rates under realistic paddock conditions.
The model contains coefficients that drive crop growth and development over time.
These coefficients cover biological processes at the heart of crop growth without making the computations too complex - aspects such as photosynthesis, or nutrient and water absorption from roots.
Genomic data is then linked to those plant-growth coefficients via advanced analytical algorithms.
The AI is then able to detect better-performing genetic combinations for specific environments and climate conditions.
The crop growth model provides a way to harness the explosion in genomic and phenotyping data in ways that are biologically meaningful and useful to growers.University of Queensland Professor Graeme Hammer.
Australia's advantage in this area arises from past Grains Research and Development Corporation (GRDC) investment in the critical component: a crop growth model within the Agricultural Production Systems sIMulator (APSIM).
The APSIM model has been adopted internationally and has continued to evolve in capability over the years.
A better way to breed
Seen as the linchpin of prediction-based crop improvement, APSIM has been used by QAAFI's Professor David Jordan to help breed high-yielding sorghum with improved adaptation to hot and water-limited environments.
The centrality of the crop growth model means that data about a plant's genotype (G) gets to interact with simulated environmental conditions (E) and management practices (M) to determine the genome's likely impact on plant growth, stress resilience and yield.
This three-way interaction is what occurs in the paddock, where it is known as the genotype-by-environment-by management interaction (GxExM).
It is this filtering, weighting and refocusing of genomic data through a GxExM prism that lifts the predictions to unprecedented and functionally useful levels.
In contrast, existing predictive technologies - such as statistical quantitative genetics and genomic prediction - attempt to make predictions about crop performance based directly on genetic data.
These kinds of genetic-based predictions were seen to falter when GxExM interactions are essential - which is the case for most of Australia's cropping systems, Professor Hammer says.
"Genomic technologies can be used in clever and useful ways, but it was naive to think that they alone could form the basis for predictions about plant performance in a paddock," he says.
It is that limitation that the new way of making predictions is overcoming. For example, the crop growth model views variation in traits, such as tillering or flowering time, in terms of combined impacts on the size of a canopy and its use of water and light.
It can predict that a genotype that flowers early and doesn't tiller much is great in a drought, but has low yield potential in a good season. Other predictive tools can be blind to such interactions.
Genomic technologies can be used in clever and useful ways, but it was naive to think that they alone could form the basis for predictions about plant performance in a paddock
, University of Queensland Professor Graeme Hammer said.
"The new approach makes it possible to select breeding material in terms of the right mix of interacting traits for specific environments," Professor Hammer says.
In the process, a new level of genetic gain in crop breeding becomes possible.
Firstly, a multitude of potential breeding material can be virtually tested in a fraction of the time required by field trials. This vastly accelerates selection of trait combinations best suited to different environments.
In addition, the same platform can be used to design the best breeding strategy to make the most useful genetic gain, including tolerances to environmental stresses.
The new approach makes it possible to select breeding material in terms of the right mix of interacting traits for specific environments.
- University of Queensland Professor Graeme Hammer.
Commercial implications
APSIM has also driven private sector developments, most notably through a past partnership between Professor Hammer's UQ team and Pioneer Hi-Bred International.
That partnership ultimately led to the development of popular and high-yielding maize hybrids (branded as AQUAmax) with improved tolerance to drought stress.
The AQUAmax varieties were developed by UQ alumni Professor Mark Cooper, who returned to the university in 2018 to occupy a chair partly supported by GRDC investment.
Part of Professor Cooper's role is to help increase public and private sector R&D collaboration.
The aim is to use crop-growth prediction to select the most suitable breeding lines and crosses, then follow this up with accelerated breeding techniques, such as 'speed breeding'.- University of Queensland Professor Graeme Hammer
"The main issue between private and public versions of this technology is scale," Professor Hammer says.
"Pioneer has a gigantic database of field trial and genomic data to help train an AI to detect the best-performing broad-scale genomic patterns.
"They interfaced this with crop modelling to target genomic predictions at paddock-relevant phenotypes. The approach is novel and it works."
Now, UQ is developing something similar for Australian pre-breeding programs, starting with sorghum and expanding to other crops, including work on wheat by UQ's Dr Lee Hickey.
The aim is to use crop-growth prediction to select the most suitable breeding lines and crosses, then follow this up with accelerated breeding techniques, such as 'speed breeding', also developed by Dr Hickey.
Field trials will still be needed to help validate the AI-based predictions.
Climate influence
Professor Hammer says climate data is a key driver of the crop growth model.
This allows the new breeding system to preview likely seasonal conditions and allow for the selection of traits that might be able to lessen the effects of climate change.
As the temperature increases, the atmosphere becomes drier and the associated increased vapour pressure deficit sucks more moisture from crops. That forces plants to use more water to achieve the same growth.University of Queensland Professor Graeme Hammer.
"I originally started to work on APSIM because I thought it was the only way we could get a handle on the climate variability that Australian growers face," Professor Hammer says.
"The crop growth simulation studies started out by running historical seasonal conditions for the past 100 years to test possibilities to improve a farm's operation and financial resilience, with regards to rotations, crop and variety choices and management practices."
Since then, climate change forecast models have improved enough that the prediction-based crop improvement tool can now be used to test germplasm against future climate scenarios.
Professor Hammer considers climate forecasts are at their most robust when it comes to temperature predictions. Already the climate has heated by an average of 1˚C. The forecasts indicate a further 1˚C (plus or minus 0.5˚C) rise in the next 30 years.
As the temperature increases, the atmosphere becomes drier and the associated increased vapour pressure deficit sucks more moisture from crops. That forces plants to use more water to achieve the same growth.
There is an upside, however, in that higher atmospheric carbon dioxide levels enable plants such as sorghum to use water more efficiently.
The problem Professor Hammer has identified is that by 2050 the extra heat is predicted to take over from the enhanced water-use efficiency associated with the carbon dioxide increase.
This finding has motivated UQ researchers to target heat tolerance, as seen with the sorghum breeding work.
Ultimately, however, Professor Hammer says that temperature increases will inevitably negate the gain.
"It means we are fighting just to retain existing yield potentials, with our gains being swallowed up by warmer, drier conditions."
This is why he sees public and private sector research collaborations as being so important, because they increase the research resources and efficiency needed to try to make up for the world's slow response to climate change.
Contact: Professor Graeme Hammer, T. +61 (0) 7 3346 9463, E. g.hammer@uq.edu.au or Carolyn Martin QAAFI Communications M. +61 (0) 439 399 886 or carolyn.martin@uq.edu.au
Original article published 8 April 2019 in Ground Cover by Gio Braidotti, New breeding technology draws on genomics, paddock realities and computer power.
