ARPA-H IGoR Wants to Make Biomedical Discovery 10x Faster — With AI Running the Lab

Most biomedical breakthroughs still happen the way they did in the 1990s: a researcher reads a paper, forms a hypothesis, spends months designing an experiment, runs it in a single lab, writes up the results, and waits for peer review. The process works — eventually. But for diseases with multifactorial causes, poorly understood mechanisms, and dynamic biological interactions across body systems, "eventually" means decades of fragmented effort while patients wait.

On May 5, ARPA-H announced IGoR — the Intelligent Generator of Research — a program designed to compress that timeline by an order of magnitude. The premise is ambitious even by ARPA-H standards: build an AI-powered ecosystem that can autonomously identify knowledge gaps in disease biology, design the optimal experiments to close those gaps, execute them across a distributed network of validated laboratories, and feed the results back into continuously improving disease models. Not as a concept study. As working infrastructure.

Solution summaries are due June 25, 2026. Full proposals follow on August 6. If you work at the intersection of computational biology, AI/ML systems, laboratory automation, or disease research, this is the most consequential open solicitation in your field right now.

The Problem IGoR Is Built to Solve

Biomedical research is fragmented by design. University labs specialize. Funding flows through siloed institutes. A materials scientist studying protein aggregation in Alzheimer's disease may never see the work of a computational biologist modeling the same pathway three buildings away — let alone collaborate with a clinical research organization running a parallel trial. Scientific insights travel slowly between fields. Valuable findings remain buried in papers that the right researchers never read.

This fragmentation is especially devastating for complex diseases. Neurodegenerative disorders, autoimmune conditions, chronic pain syndromes, and emerging infectious diseases don't respect disciplinary boundaries. They involve cascading interactions across molecular, cellular, tissue, and organ-system scales. Understanding them requires integrating data from genomics, proteomics, imaging, clinical phenotyping, and environmental exposure studies — data that currently lives in incompatible formats across thousands of independent laboratories.

The result is what ARPA-H program manager Paul E. Sheehan, Ph.D. describes as a research ecosystem where "discoveries remain isolated across different labs and disciplines," coordination failures erode research credibility when results cannot be verified, and the current discovery process relies heavily on chance connections between researchers.

IGoR is ARPA-H's attempt to replace chance with architecture.

Four Technical Areas, One Closed Loop

IGoR's design is structured around four interdependent technical areas that, together, form a continuously cycling research engine.

Technical Area 1: Mechanistic Disease Models. Teams must build computational models that encode causal biological relationships across molecular, cellular, and physiological scales. These aren't statistical black boxes — they're mechanistic representations of how diseases actually function. The models must be detailed enough to generate testable hypotheses and flexible enough to incorporate new experimental data as it arrives. Think of them as living biological knowledge graphs that get sharper with every experiment.

Technical Area 2: AI Orchestration Layer. This is the brain of the system — an AI engine that analyzes the disease models, identifies where knowledge is missing or uncertain, and designs the specific experiments most likely to fill those gaps. The system must prioritize experiments by expected information gain, not just feasibility, and it must reason across disciplines. If a protein-folding question requires both cryo-EM structural data and transcriptomic expression profiles from patient tissue, the AI should design both experiments as a coordinated package.

Technical Area 3: Protocol Architecture. This may be the most quietly transformative component. IGoR requires a layered protocol system that separates scientific intent from hardware-specific execution instructions. The goal is to make experimental procedures transferable between laboratories with the same ease as digital data transfer. A protocol designed at MIT should execute identically at a contract research organization in San Diego or a national laboratory in Tennessee — same scientific intent, adapted to local equipment, verified for reproducibility. This is the infrastructure layer that makes the distributed marketplace possible.

Technical Area 4: Distributed Laboratory Marketplace. ARPA-H envisions a network of validated, qualified laboratories that can execute standardized protocols and return gold-standard data. Labs in this marketplace would operate somewhat like cloud computing nodes — accepting experimental workloads, executing them according to certified protocols, and returning verified results. The marketplace creates competition on quality and turnaround time while ensuring reproducibility through the protocol architecture.

When these four components work together, they form a closed loop: models generate hypotheses → AI designs experiments → protocols standardize execution → labs run experiments → data flows back into models → models improve → the cycle accelerates.

Who Should Apply — and Who Shouldn't

ARPA-H anticipates awarding approximately three multidisciplinary performer teams, each of which must address all four technical areas. The awards will be structured as Other Transaction (OT) agreements — not traditional grants — giving ARPA-H more flexibility in structuring milestones and deliverables.

The required expertise spans an unusually wide range: computational biology and mechanistic modeling, AI/ML orchestration and agentic systems, laboratory automation and robotics, experimental protocol standardization, distributed systems architecture, and human-centered interface design. Teams must include capabilities in disease biology, data engineering, and validated wet-lab experimentation across multiple modalities.

This is not a program for a single PI with a good idea. IGoR demands industrial-scale integration across disciplines that rarely work together. The most competitive proposals will likely come from consortia that combine academic disease expertise with technology companies that have built large-scale AI and automation systems. If your team cannot credibly address all four technical areas, the solution summary is unlikely to advance.

The disease focus is intentionally broad: neurodegenerative disorders, autoimmune diseases, chronic pain syndromes, and emerging infectious diseases. ARPA-H is not prescribing which disease to model first — they want teams that can demonstrate the architecture works and then extend it.

The Timeline: Five Years, Three Phases

IGoR is structured across three phases over five years:

Phase I (18 months): Concept and component development. Teams build individual technical area capabilities, demonstrate proof-of-concept integration, and establish baseline performance metrics. This is where the disease models get built, the AI orchestration engine gets trained, the protocol architecture gets designed, and initial laboratory partnerships get formed.

Phase II (18 months): Cross-team integration and interoperability. The three performer teams must demonstrate that their systems can interoperate — sharing protocols, exchanging data, and validating each other's results. This is the phase where IGoR becomes an ecosystem rather than three parallel projects. ARPA-H will likely push for standardized interfaces that allow components from different teams to work together.

Phase III (24 months): Scaling, transition, and commercialization planning. The integrated system must demonstrate the 10x acceleration target on real disease research questions. Teams must develop transition plans for how the infrastructure persists after ARPA-H funding ends — whether through commercialization, adoption by NIH or other agencies, or open-source community governance.

A Proposers' Day event is planned for the Washington, D.C. metro area (details forthcoming), and ARPA-H has launched a teaming page for organizations looking to find collaborators.

What Makes IGoR Different From Every Other AI-in-Biology Initiative

The last two years have seen a surge of AI-biology programs: NIH's Bridge2AI, DOE's GeneSIS, Google DeepMind's protein structure predictions. IGoR is distinct in three ways.

First, it's not just about AI analysis of existing data. IGoR includes the physical laboratory infrastructure to generate new data. The distributed marketplace of validated labs means the AI doesn't just suggest experiments — it can orchestrate their execution and receive the results.

Second, it requires mechanistic models, not correlative ones. While most AI-biology efforts train neural networks on large datasets to find statistical patterns, IGoR demands causal models that encode how biological systems actually work. This is harder to build but produces explanations that scientists can interrogate, validate, and extend.

Third, the protocol architecture addresses reproducibility by design. The replication crisis in biomedical research — where up to 70% of published results cannot be independently verified — is partly a protocol standardization failure. IGoR's requirement for transferable, hardware-agnostic protocols could create infrastructure that outlasts the program itself.

A Note on ARPA-H's Broader Momentum

IGoR joins an increasingly crowded portfolio of ARPA-H programs that collectively signal the agency's ambition to rewire — not just fund — biomedical research. The $139 million EVIDENT initiative is building clinical evidence pipelines for behavioral health therapeutics including psychedelics. ADVOCATE is deploying agentic AI for cardiovascular care. STOMP is tracking microplastics through the human body. PROSPR is funding $144 million in aging therapeutics.

The pattern is clear: ARPA-H is not distributing R01-style grants for incremental advances. It's building research infrastructure at a scale and speed that no other federal agency is attempting.

What to Do Before June 25

The IGoR solicitation (Notice ID: ARPA-H-SOL-26-155) is posted on SAM.gov and accessible through the ARPA-H Solutions Portal. Solution summaries are a required step — ARPA-H will use them to encourage or discourage full proposal submissions.

If you're a computational biologist, AI researcher, or laboratory automation company considering this program, the teaming page is where you start. No single organization has all four technical areas covered. The winning teams will be consortia that have already identified their partners by the time they submit the solution summary.

And if the scope feels beyond your reach but you're working in disease biology, protocol standardization, or lab automation — watch the IGoR teaming page for organizations actively recruiting collaborators. The best way to participate may be as a specialized partner within a larger team.

For organizations tracking ARPA-H opportunities across programs, tools like Granted can surface solicitations as they post, match them to your capabilities, and help you draft competitive proposals before the deadline clock starts running.