ORNL Expands OPAL Agentic AI for Nickel Phytomining Experiments

ORNL reports a new OPAL phytomining experiment, not a new OPAL launch

Oak Ridge National Laboratory published new results on August 28, 2026 from its OPAL agentic-AI program for autonomous science. The new development is a nickel-uptake experiment on pennycress and a substantially expanded set of automated co-scientist workflows. ORNL says OPAL itself had already demonstrated faster plant-data analysis before this experiment, so this should not be described as the first launch of the platform.

The work is part of the Department of Energy's multi-laboratory Orchestrated Platform for Autonomous Laboratories (OPAL) effort, involving Oak Ridge, Argonne, Lawrence Berkeley and Pacific Northwest national laboratories. ORNL's current focus uses AI, robotics and automated experimentation to accelerate biological approaches for recovering critical minerals.

The experiment tracked 360 plants and more than 24,000 observations

Researchers tested 12 distinct pennycress lines collected from different growing regions. A total of 360 plants were grown in soils with varying nickel concentrations inside ORNL's Advanced Plant Phenotyping Laboratory.

Automated imaging repeatedly measured changes in leaf color, plant architecture, growth, stress response and mineral content. By the end of the experiment, ORNL says the system had generated more than 24,000 observations across the plants.

The resulting data were analyzed with deep-learning methods and the OPAL co-scientist, which plans analyses, writes and runs code through tools, retrieves results, supplies context and proposes next steps for researchers.

Expanded OPAL workflows surfaced nickel-tolerance signals

ORNL says the latest co-scientist workflows identified the top nickel-tolerant pennycress varieties in the experiment, assessed which plants showed stronger growth and stress resilience, and found early-stage traits that best predicted final nickel accumulation.

The system also flagged anomalies for researchers to inspect and suggested follow-on experiments intended to validate growth dynamics, investigate nickel-tolerance biology and refine future biodesign strategies.

These are results from a controlled research experiment. They do not establish that pennycress phytomining is already deployed commercially at scale or that the identified plant lines will deliver the same performance under field conditions.

More than 1,000 plant traits can be queried in minutes

ORNL reports that the co-scientist reduced analysis time for more than 1,000 physical plant traits from hundreds of hours of manual work to a few minutes of interaction.

In a manual comparison during the experiment, two researchers spent about six hours recording 10 plant traits at one timepoint. ORNL says the automated platform can reproduce the hand-collected measurements in less than a minute while also extracting and analyzing hundreds of traits from repeated imaging.

Those comparisons describe ORNL's experimental workflow and should not be treated as a general benchmark for every autonomous laboratory, biological dataset or AI agent.

The system is now connected to broader DOE science infrastructure

ORNL says the team demonstrated transfer of optimized AI-ready data to the American Science Cloud and a Department of Energy biological and environmental research database. The OPAL co-scientist was also deployed within the American Science Cloud.

That infrastructure connection is significant because autonomous-science systems become more useful when experiments, models, agents and datasets can be shared across laboratories instead of remaining isolated within a single facility.

What is released now versus still upcoming

Demonstrated now: the pennycress nickel experiment, expanded automated OPAL workflows, analysis of more than 24,000 observations, identification of nickel-tolerance and predictive traits, anomaly detection, follow-on experiment suggestions, and deployment of the co-scientist within the American Science Cloud.

Still upcoming: ORNL plans to combine the plant experiments with proteins and microbes being developed by OPAL partners. Argonne is working on proteins for bioleaching, while Pacific Northwest and Lawrence Berkeley are developing metal-tolerant microbes using robotics and agentic-AI approaches.

The broader goal is a programmable phytomining system that co-engineers plants, microbes and proteins. That multi-component system remains a research direction rather than a completed commercial deployment.