Argonne National Laboratory is building AI systems that design experiments, reason about biology, and run robotic labs with minimal human help. Three new projects, funded through the Department of Energy's Genesis Mission, aim to compress years of biological research into weeks by pairing machine learning with self-driving laboratories - automated facilities where algorithms and robots handle the bench work.
The Genesis Mission is a DOE initiative to put AI at the center of scientific discovery. Argonne brings a rare combination of assets to the effort: world-class computing, deep AI expertise, and hands-on biology labs that already use robots to run experiments. The three projects - OPAL, IDeA, and MELT-REE - each attack a different biological problem, but they share a common blueprint: let AI and automation do the repetitive work so scientists can focus on what the data means.
OPAL: a network of self-driving labs
OPAL, short for Orchestrated Platform for Autonomous Laboratories to Accelerate AI-Driven BioDesign, connects self-driving labs across four national laboratories. Argonne handles protein design. Lawrence Berkeley National Laboratory and Pacific Northwest National Laboratory study microbial behavior. Oak Ridge National Laboratory focuses on plant biology. The goal is a system where an AI planning agent designs an experiment, sends tasks to the right lab, watches the results come in, and adjusts the next round - all without a human pushing buttons.
Argonne's edge in the project includes humanoid robotics. Most lab automation uses liquid-handling machines, but some experiments need a robot that can adjust instrument settings, read visual cues, or operate equipment built for human hands. "There are some experiments for which we need robots that can do specific types of experiments," said Arvind Ramanathan, a computational biologist at Argonne. "We're pushing on that. We have humanoid robots in the lab where we are experimenting with what they can do and how we train them."
Running experiments across institutions simultaneously - where a single workflow starts at Argonne, moves to Berkeley Lab, and scales across both - demands software and coordination infrastructure that does not exist off the shelf. Building that system is part of the science OPAL is doing.
IDeA: AI agents that search, reason, and debate
Characterizing a new enzyme can take a trained team a year or more. The Intelligent Design Assistant for Enzyme Discovery and Biosynthetic Pathway Optimization, or IDeA, wants to shrink that to weeks. Led by Ramanathan, IDeA deploys AI agents - specialized digital researchers - that comb through roughly 3 million scientific papers, scan biological databases, compare molecular structures, and generate hypotheses in parallel on a supercomputer.
The first target is enzymes that produce nylon-like biopolymers. Finding and optimizing the right enzyme today means months of literature review, database searches, and lab testing spread across a research team. IDeA automates that workflow and runs many lines of inquiry at once. The project also tackles a harder problem: building AI that reasons about biology rather than just retrieving facts. Argonne trains biological reasoning models on data generated in its own labs. When multiple AI agents disagree, the system is designed to resolve the conflict and stay grounded in established science.
"Argonne hosts some of the best bioinformatics data sets on the planet," Ramanathan said. Oak Ridge National Laboratory serves as the experimental partner, generating biological data that feeds back into the AI models - closing the loop between computation and the bench. For scientists looking to build these skills, structured training in AI for Research Scientists covers the data modeling and experimental design techniques that underpin systems like IDeA.
MELT-REE: microbes that mine rare-earth elements
Rare-earth elements are essential to electronics and defense systems. The U.S. has deposits, but extracting them is expensive and chemically harsh. MELT-REE - Multimodal Engineering and Leaching Technology for Rare-Earth Extraction - tests whether bacteria can do the job instead. The process, called bioleaching, uses microbes to pull rare-earth elements from mine tailings and electronic waste. It works at small scales. Making it fast and efficient enough for industry is the hard part.
"Conventional chemical leaching works, but it comes at a cost - corrosive reagents, significant energy input and acid waste that has to go somewhere," said Daniel Schabacker, an Argonne bioscientist who co-leads the project with Ramanathan. "Biology offers a fundamentally different approach: a microbe that does the same job without the chemical footprint."
The bacteria generate acid as they leach, but that same acid eventually inhibits them. They also struggle when metal concentrations rise or when solid feedstock exceeds about 1% by weight - industrial processes need 10% or higher. Schabacker's team screens a library of 2,733 bacterial strains, each with a different gene knocked out, using Argonne's self-driving lab. The data goes straight to Ramanathan's computational team, which uses AI to spot patterns and predict which genetic changes will produce a tougher, faster microbe. This tight integration of automated experimentation and AI analysis - part of the broader shift toward AI for Science & Research - is what makes the work possible at Argonne.
Why this matters for science and research professionals
These projects signal a structural shift in how biological research gets funded and executed. Self-driving labs and AI reasoning agents are moving from proof-of-concept to multi-institutional infrastructure. For researchers, that means experimental design skills and data literacy will matter as much as bench technique. The laboratories that win funding and produce results fastest will be those that can connect computation, automation, and domain expertise - exactly the model Argonne is building with OPAL, IDeA, and MELT-REE.
Your membership also unlocks:
