b'at play, this prediction becomes a challeng- approximately five years. the preparatory phase, we have already been ing if not daunting task. CropXR pursuesSubsequently, these models can beusing existing data sets. We are summariz-the creation of new plant models that willtranslated to specific crops. When thising and analyzing them and using them to provide insight into how the interplay ofphase arrives, we will need significantly lessfuel our first model versions. multiple genes together shapes plant resil- crop-specific data, since the primary modelWe are keen on using high-quality ience. Instead of examining how one singleis already available, and it only needs to bedata for our models since the accuracy of stressor impacts a plant, we look at the effectadapted to a particular crop. This will makeour models and predictions depends on of multiple stresses. They can be biotic, forit easier to create so-called spin-off modelsthis. Our team of data scientists, led by a instance, stress caused by bacteria. Or theyfor other crops. Breeders can then use thesededicated technology director, is working can be abiotic, for instance, temperature.models to step up their game.on creating a robust data infrastructure to Important questions include: does this leadstore, manage and access all this data prop-to trade-offs, which gene combinations areSWE: AI IS AN IMPORTANT ELEMENTerly and safely. It will be based on the FAIR key determinants of resilience, and whatOF CROPXR. CAN YOU TELL US MOREprinciples (Fair, Accessible, Interoperable, outcomes can be expected? These questionsABOUT THE USE OF AI WITHIN YOURReusable). We intend to create an inclusive relate to the breeding targets of the future. CONSORTIUM? HOW DO YOU HANDLEdata culture. All those within the consor-CropXR finds itself on the verge of aTHESE ENORMOUS SETS OF DATA ANDtium should be able to access the data. new era in plant science in which the aim isWHO CAN ACCESS THEM? This solid foundation will lead the way to know and predict the intricate dynamicsHT: The size of the data sets is indeed quiteto realize one of our ambitions, which is to of complex traits and genes. To obtain theseoverwhelming. Our data scientists estimatedevelop a Resilience Hub. This hub should insights, new state-of-the-art methodolo- that our project will create a data volume ofbe an extensive collection of knowledge on gies are necessary.approximately several petabytes in the nextcrop resilience that will corral knowledge, We deploy, therefore, the method- few years. However, bear in mind that thisdata, tools and experts related to this sub-ology of smart data breeding. This meansvolume is relatively modest compared to theject. It will be a treasure trove for all those that we collect a large volume of measure- data used in, for instance, medical sciencesworking in this field. We strive to realize this ments and combine various types of dataor astronomy.ambition in the years to come. The require-sets to construct our models. First, we try toMachine learning can help us to sum- ments and possibilities to access the hub by construct these models based on the earli- marize all the experimental data and usethose outside the consortium will also be fur-er-mentioned plant Arabidopsis. Compilingthem for our models. Soon, we will com- ther developed in the coming years. We have these models is an iterative process that willmence our first experiment on droughtonly just started, and a lot of work needs to be take time. We expect to be able to modeland temperature stress in Arabidopsis. Thedone first. We are eager to keep all those in Arabidopsis in the next couple of years,experiment is currently being prepared andthe seed sector informed on our progress and leading to a more refined model afterwill run for approximately a year. Duringdeliver cutting-edge results in due time! SEEDWORLD.COM/EUROPEISEED WORLD EUROPE I 21'