Artificial intelligence to understand plant resilience in harsh environments

The Atacama Desert, located in South America, is one of the driest regions on Earth. Several types of endemic plants are still present at the site. After collecting several species that grow between 2,400 and 4,500 meters above sea level, scientists from INRAE, Purdue University and the Pontifical Catholic University of Santiago in Chile have been able to identify common molecular markers that allow an understanding of the mechanisms of these plants’ resilience in the face of a harsh environment. The researchers used an innovative approach using artificial intelligence. The results of their work are detailed in review The new botany.

“We used a technique called predictive metabolism, Explains Pierre Petrick, Research Professor at Purdue University and INRAE. It consists in the study of a metabolite, that is, all the small molecules that are present in a living organism, such as sugars, amino acids and antioxidants. Then the originality of our approach lies in the combination of this technology andartificial intelligence. »

Concretely, 24 species were analyzed and made it possible to study nearly 5,000 different metabolic signals. Thanks to the use of machine tendency, a statistical approach based on training a machine learning algorithm, scientists have succeeded in highlighting 39 molecules Common to all these species which makes it possible to predict the environment in which the plant grows with an accuracy of 79%. Among these 39 molecules,starch It has been identified as the best predictive chemical compound.

“The prediction rate is very high, we did not expect it, researcher analyzes. With phenotypic and physiological data such as plant size, productivity or photosynthesis, such high resolution cannot be achieved. In addition, our study was conducted on plants of several species, while researchers are often only interested in one species in particular. » From this result, the researchers were able to correlate the expression of these 39 molecules with various environmental stress factors such as water stress, very high light intensity, high soil salinity, etc.

Reduce development times for new plant species

Since these 39 chemical compounds are also found in plant species grown in other climates, such as corn, peas, tomatoes or sunflowers, this new approach could also help to better target plants whose metabolic processes are more likely to withstand environmental stresses. . “This is a research hypothesis that needs to be tested, Says Pierre Pétriacq. The idea would be, for example, to find faster varieties whose metabolism is close to that observed in plants of the Atacama Desert. This can reduce the time needed to adapt cultivated plants to the constraints of climate change. Today, it takes scientists and producers ten years to develop new plant varieties that meet specific environmental criteria such as frost resistance. »

Another means of application, this time at the agricultural level. By using predictive metabolism that combines the study of chemical composition and the use of artificial intelligence, the researchers plan to try to predict the behavior of plants according to different farming practices. “We can imagine scenarios where we grow plants with fairly benign practices, such as reducing phytosanitary products, and then try, for example, to identify species associations that are part of an agro-ecological approach.”The researcher adds.

Finally, the last line of action can apply not to our planet, but to Mars. The Atacama soil is actually the soil most similar to that of Mars. Researchers on this project are in contact with scientists from the European Space Agency and NASA and have embarked on a reflection that may lead them to use this new approach to identify plants likely to grow on Mars. Of course, the idea is not to try to grow it on the planet, but to do basic research work.