Crag News

GenoDrawing: A revolutionary AI tool for predicting complex traits in plants

A new tool called GenoDrawing can predict the shape of apples from their genetic markers using deep learning
  • The software can be used to analyse complex traits in plants, which are difficult to measure with traditional methods.
  • This study demonstrates the potential of AI for advancing plant breeding and agriculture.

A team of researchers led by Maria José Aranzana, IRTA researcher at CRAG, has developed a groundbreaking tool based on generative AI for predicting complex traits in plants. The tool, called GenoDrawing, uses autoencoders and deep learning approaches to predict apple images from genetic markers known as single nucleotide polymorphisms (SNP), opening new prospects for plant breeding.

The study, published in the journal Plant Phenomics, introduces a new methodology that predicts physical traits of an apple variety from its genetic information, generating images that closely resemble reality. The researchers used a deep learning approach to train the GenoDrawing model on a dataset of more than 10,000 apple images and their genetic information (SNP markers), and then tested its efficiency on a separate dataset. The results showed that GenoDrawing was able to predict the shape of apples and reproduce it in images based solely on the genetic information coming from SNP markers with a high degree of accuracy.

One of the advantages of GenoDrawing is that it can be used to predict a complex trait captured through images, such as shape, and with the potential of predicting other characteristics, such as fruit colour, leaf morphology, or plant architecture, in further studies. This is important because traditional methods of plant phenotyping, such as visual inspection, are time-consuming and expensive and may not provide accurate results. Besides, with GenoDrawing, researchers can quickly and easily foresee the traits of plants based on their DNA without having to grow the plants or wait for them to bear fruit, which could lead to more efficient and effective breeding programs.

The researchers emphasize that the power of GenoDrawing relies in the utilization of artificial intelligence (AI):

"Our study shows that AI can be used in new ways to link visual and genetic information thus predicting complex traits in plants" states Federico Jurado-Ruiz, the first author of the published manuscript.

Dr. Aranzana is confident in the potency of the tool:

"We believe that this tool could have a wide range of applications, not only in plant breeding and agriculture but also in different fields of life research, and we look forward to exploring its potential in future research."

The use of generative AI in GenoDrawing tool represents a major step forward in the field of plant phenotyping and breeding. By harnessing the power of deep learning approaches, researchers can now predict complex traits in plants, such as shape of apples which could have significant implications in plant breeding and agriculture. The researchers hope that their work will inspire further research into the use of deep learning approaches for predicting complex traits, and that it will lead to new breakthroughs in the field of agriculture.

Reference article

Jurado-Ruiz F, Rousseau D, Botía JA, Aranzana MJ. GenoDrawing: An Autoencoder Framework for Image Prediction from SNP Markers. Plant Phenomics 2023;5:Article 0113. https://doi.org/10.34133/plantphenomics.0113

About the authors and funding of the study

Federico Jurado-Ruiz is recipient of grant PRE2019-087427 funded by MCIN/AEI/10.13039/501100011033 and by "ESF Investing in your future". This research has been supported by project PID2021-128885OB-I00 funded by MCIN/AEI/10.13039/501100011033 and by "ERDF A way of making Europe". This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement no. 817970 (INVITE). This work was also supported by the CERCA Programme ("Generalitat de Catalunya"), and by the "Severo Ochoa Programme for Centres of Excellence in R&D" 2016-2019 (SEV-2015-0533) and 2020-2023 (CEX2019-000902-S) both funded by MCIN/AEI/10.13039/501100011033.