Phenotyper - One who measures biological systems because you can't understand what you don't measure.
Understanding systems of plants, microbes, and soils requires diverse and tedious measurements and frameworks for connecting all the information. My work leverages diverse measurement modalities from gas flux to imaging to make sense of the complicated whole.
My past work has led me to a convergence of AI, real-time sensing, connectivity, and automation that will allow autonomous discovery across science enabled by ORNL's big team science opportunities. Autonomous discovery relies on self-driving labs and AI to determine next experiments, collect data, and analyze with limited human intervention, with the ultimate goal of benefitting research and augmenting the work force. I am currently developing soil incubations with real time measures and analysis, a robotic system to image petri dishes, and a real-time dashboard leveraging the ORNL INTERSECT microservices framework. An AI collaborator can interpret visualized plots, provide suggestions, and discuss with the user via a chat interface. While self-driving labs for pharmaceuticals and microbial synthetic biology are now widespread, automating plant, microbe, and soil research requires a step towards more general purpose robotics rather than liquid and plate handling. While the biological applications are discussed below, I am driven to create: the experiment you can talk to.
Springing from my fascination with plant roots beginning with root behavior and game theory, I have studied whole-plant ecophysiology with a focus on roots for 20 years. Quickly, I adopted a microeconomic view of plant form and function focused on carbon allocations and expenditures in my first paper to helping develop the root economics space. To measure the hidden dimensions of form and function, I develop root and shoot phenotyping platforms ranging from flux-based to image-based, over time leading to automated workflows and AI approaches. I have leveraged 15N-stable isotope studies for demonstrating root activity, and am currently using new 13C flux instruments. At ORNL, my work has integrated with research areas like Center for Bioenergy Innovation with a focus on improving plant productivity for refining sustainable aviation fuel and Plant Microbe Interfaces SFA focused on understanding the poplar microbiome and how systems tolerate stress. My current biological aims are to understand carbon balance among organisms within the plant-microbe-soil system and how to alleviate tradeoffs to drive performance.
My organizational missions have at times focused me on fundamental research on plant form and function, as well as more applied trait discovery for direct breeding pipelines. I proposed functional phenomics as a research strategy for understanding plant physiology using high-throughput phenotyping of large diversity panels, bioinformatics, and simulations. To bring root traits into breeding pipelines and genetic studies, I developed RhizoVision Explorer as the number one root image analysis software with more than 14,000 downloads, imaging platforms for field root crowns, greenhouse gantry shoot imaging, a high-throughput root nutrient uptake platform, and root respiration. I have leveraged image-based methods for 10 years, and now lead image and data analysis for the Advanced Plant Phenotyping Laboratory and am increasing focus on hyperspectral imaging and AI methods, specifically vision transformers. Our exascale vision transformer team has published new methods for training hyperspectral models and is currently scaling to millions of images on the Frontier supercomputer.
