Abstract:

The future of sustainable agriculture hinges on expanding crop diversity with locally adapted species. While nature harbors thousands of wild plants with untapped agronomic potential, their domestication remains slow and labor-intensive. Chenopodium album, a nutrient-dense wild relative of quinoa and spinach, emerges as a compelling candidate for crop development in temperate regions like Denmark. Its high protein content, historical consumption in prehistoric Denmark, and potential climate resilience make it ideal for domestication. However, its cultivation is hindered by bitter and anti-nutritional saponins and poorly characterized growth and yield— limiting its agricultural viability. While Machine Learning (ML)-driven breeding has advanced major crops, wild species like C. album are data-scarce. To overcome this, I will leverage data from related species (quinoa) and major crops (maize), including multi-omics and phenotypic datasets for training. 
 
PLANTAI will bridge this gap by developing a data-driven breeding framework leveraging resources generated by the host research group, including a unique C. album collection from Denmark for testing. The project has three objectives: 1) identify genetic loci and regulatory networks controlling key domestication traits using multi-omics analysis and comparative genomics, 2) develop deep learning (DL) models integrating multi-omics and phenotypic data to predict genetic gains, and 3) implement Bayesian optimization (BO) to design optimal breeding strategies, balancing yield potential and reduced saponin content. These DL models will be trained on external datasets and fine-tuned using newly generated C. album data. 
 
DL models will extract trait-informative features from multi-omics data, while BO will optimize plant crossing designs, reducing reliance on traditional breeding cycles. This AI-powered approach will accelerate C. album domestication, potentially serving as a model for other wild species. 
 
PLANTAI aligns with the integration of data science in sustainable agriculture. Beyond C. album, this framework will establish a scalable AI-driven breeding model adaptable to other wild plant species, enhancing global food security and crop resilience in a changing agricultural landscape. The project’s innovative approach combines cutting-edge genomics, ML, and optimization techniques to accelerate crop domestication, addressing urgent needs in sustainable agriculture and food production.