EcoBOT: an AI/ML enabled automated phenotyping capability for model plants

Peter Andeer^1*Petrus Zwart¹Daniela Ushizima^1,2,3Marcus Noack¹Lloyd Cornmesser^1,4Thomas Vess⁵Zineb Sordo¹Stephen Tan¹Joseph Zorzi^1,4Chelsea Hernandez¹Vlastimil Novak¹Yezhang Ding¹John Vogel^1,4Benjamin Paul Bowen^1,4James Sethian^1,6Trent Russell Northen^1,4*

• ¹Berkeley Lab (DOE), Berkeley, United States
• ²Bakar Computational Health Sciences Institute, Medical Center, University of California, San Francisco, San Francisco, California, United States
• ³Institute for Data Science, Division of Computing, Data Science and Society, University of California, Berkeley, Berkeley, California, United States
• ⁴Joint Genome Institute, Berkeley Lab (DOE), Berkeley, California, United States
• ⁵Veracyte, South San Francisco, California, United States
• ⁶Department of Mathematics, Berkeley City College, University of California, Berkeley, Berkeley, California, United States

Advances in automation and AI/ML provide promising new capabilities for plant science including design, modeling, and analysis. We describe the EcoBOT, an automated platform for researching small model plants under axenic conditions and its integration with AI/ML tools. Plants are grown on the EcoBOT within sterile containers (EcoFABs) and plant growth and health are monitored through imaging. Brachypodium distachyon grown in the EcoBOT maintained sterility and responded to nutrient limitation and copper stress. By analyzing 6,500 + root and shoot images we found that root and shoot responses to copper varied in sensitivity and response rates. Bayesian Optimization was used to improve model accuracies relating copper concentrations to plant biomass via sequential experiments by >30%. Future experiments relating other chemical stresses and microbial interactions could help create generalized models of plant responses to environmental factors.