Want to improve plant nutrient uptake with new machine learning and AI methods
Umeå University and SLU, the Swedish University of Agricultural Sciences, are now working on predicting the properties of proteins using new AI and machine learning methods.
- It is about solutions to efficiently improve nutrient uptake in plants, and reduce the impact on nature and the environment, says Tommy Löfstedt, Senior Lecturer at the Department of Computer Science, Umeå University.
Regina Gratz, Swedish University of Agricultural Sciences, SLU, and Tommy Löfstedt, Department of Computing Science, Umeå University, collaborate in a new and unique project.
Fertilizer use in agriculture is today a major problem worldwide. Around 100 million tonnes of nitrogen are used annually, and a large part goes directly into lakes, oceans and the environment.
There are essentially two ways to reduce pollution from eutrophication. Using more environmentally friendly sources of nitrogen and, improving the ability of plants to take up nutrients.
"In this project, we are linking these two possibilities, i.e. we are trying to improve the plants' ability to absorb the nutrients they receive through more environmentally friendly fertilizers," says Tommy Löfstedt, associate professor at the Department of Computer Science.
Environmentally friendly agriculture
It is special transport proteins that manage nutrient uptake in plants, and by changing the properties of these proteins, uptake can be improved.
"By identifying beneficial protein properties, it is possible to use them to their advantage. The problem is that such research relies on both time-consuming and costly laboratory experiments", says Tommy Löfstedt.
With funding from the Kempe Foundations, he and Regina Gratz, post-doctor at the Department of Forest Ecology and Management at SLU, the Swedish University of Agricultural Sciences, will now develop new machine learning methods that can both streamline and facilitate these experiments.
The possibilities of AI
Machine learning and AI are very briefly about training computers to automatically detect and learn a set of rules to solve a particular task. The project team will now develop such methods and test them in SLU's laboratory in Umeå.
"Our machine learning methods should be able to select the right target proteins in advance and speed up practical laboratory work. At the same time, we want to increase the probability of efficiently finding relevant protein regulations," says Tommy Löfstedt.
EU wants to reduce fertiliser use
The EU wants to reduce the use of fertilizers by 20 percent and see a 50 percent reduction in losses, while at the same time, increasing the share of organic farming by 25 percent.
Using AI methods to predict protein properties are areas of research and innovation that have been established over the last year and are changing rapidly. "However, our combination of AI and this biological question is, as far as we know, unique," says Tommy Löfstedt.
In a few years, the project team hopes to present new AI tools that will greatly reduce the time it takes to test protein properties in the lab, as well as to make these tools widely available.