"Families shouldn't wait for breakthroughs while new knowledge trickles through the literature and researchers do experiments that are the most familiar rather than the most informative," said Alicia Jackson, ARPA-H director.
ARPA-H's Igor: an end-to-end autonomous research infrastructure
The Advanced Research Projects Agency for Health (ARPA-H), a unit of the U.S. Department of Health and Human Services, has proposed a five-year program called Igor that aims to rework how biomedical research is done. ARPA-H frames the effort as a move away from research driven by individual labs toward "an AI-enabled ecosystem of shared research and experimental procedures" that it says could speed breakthroughs by as much as ten-fold.
ARPA-H emphasizes that Igor is not merely "to put a biomedical research wrapper around a frontier large language system." Instead, the agency describes an integrated cycle that will "identify promising new research, select a testable proposition, design an experiment and find labs to conduct them." The agency also says Igor will make transferring an experiment from one lab to another "as straightforward as sending a data file."
ARPA-H justifies the effort with stark reproducibility estimates: more than 70% of researchers cannot reproduce another scientist's experiments and up to 89% of preclinical work cannot be fully reproduced—gaps the agency casts as delaying treatments for chronic and complex diseases.
Four technical pillars ARPA-H laid out
- Computational disease models: ARPA-H wants models that "move beyond statistical correlation and instead represent how diseases actually function biologically" across scales from molecular and cellular interactions to tissues, organs and whole-body systems.
- AI orchestration layer: an intelligent layer that "identifies knowledge gaps and designs the optimal experiments for researchers to run."
- Layered protocol architecture: standardized protocols intended to let any qualified laboratory "execute the same experiment reproducibly."
- Distributed marketplace of validated laboratories: a network of labs that execute standardized protocols and return "gold-standard" data.
Combined, ARPA-H says, these components will create "a cycle of hypothesis generation, experimentation and model refinement that enables researchers to create validated knowledge" far more rapidly than current approaches and to empower researchers to pursue bold, unconventional directions currently too slow or resource‑intensive.
Proposal timeline, AI-assisted review, and funding disclosure
ARPA-H opened Igor to outside proposals on a compressed schedule. Solution summary proposals are being accepted until June 25; after submission, proposers "will either be encouraged or discouraged from submission of a full proposal." Full proposals are due Aug. 6.
ARPA-H also plans to use AI tools in the review process itself. The agency said Igor "is piloting secure large language model tools to assist with the initial review of solution summaries," citing anticipated significant interest in the program. ARPA-H did not publicly disclose the amount of funding planned for Igor.
Patient-advocacy and research group perspectives
Some biomedical experts view Igor as a potentially transformative accelerator when paired with other collaborative efforts. Christian Rubio, executive director at EverythingAL, a non-profit focused on applying data science to ALS care and cures, argued that Igor's greatest opportunity is to "move beyond institutional datasets capturing insights in closed loops" toward "continuously learning ecosystems that integrate longitudinal real-world information, digital biomarkers, molecular data, and clinical outcomes in ways that are actionable and reproducible."
"Ultimately, we would hope Igor accelerates discovery itself, and the creation of a more connected and collaborative biomedical ecosystem that helps turn a flywheel of innovation," Rubio said.
What this means for technologists, regulators, and laboratories
- Technologists and security teams will be asked to build and validate complex computational models, an AI orchestration layer, and secure interfaces for protocol transfer—plus the pilot secure large language model tools ARPA-H plans to use in proposal review.
- Policymakers and regulators are already working in parallel: the Food and Drug Administration said it is launching an expanded pilot program to use AI to increase the "efficiency, speed and quality of decision-making in clinical trials," with an aim toward enabling "real-time" trials—an initiative ARPA-H's Igor would intersect with if adopted broadly.
- Laboratories and procurement leaders will face new expectations to adopt standardized, layered protocols and to participate in a distributed marketplace that returns "gold-standard" data—measures ARPA-H says are necessary to move experiments reproducibly between sites.
ARPA-H has sketched an ambitious engineering and organizational plan: technical pillars, a marketplace of validated labs, and AI-assisted review. The agency's next concrete moves are procedural—solution summaries due June 25, full proposals on Aug. 6—and a pilot using secure large language models to triage submissions. What remains undisclosed is the program's scale: ARPA-H has not revealed planned funding for Igor. Whether the initiative will make experiment transfer "as straightforward as sending a data file" will be tested first by who responds to the June and August deadlines and by the robustness of the protocols and models those teams propose.
