ARPA-H's BIOGAMI Award Went to an Eight-Institution Consortium Assembled Before the Solicitation Posted. That's the Lesson.

On October 9, 2026, ARPA-H announced that its BIOGAMI program had begun work. The solicitation that produced it closed on April 15, 2026. For anyone still trying to win an ARPA-H Other Transaction, the award is now more useful than the solicitation ever was — because it shows what the agency actually bought.

BIOGAMI — BIOmolecular Grammar for protein Aggregation Modulation and Intervention, solicitation ARPA-H-SOL-26-149, program manager Shannon Greene, Ph.D. — targets protein misfolding and aggregation at its earliest molecular stages, before the aggregates that define Alzheimer's disease and ALS have formed. Greene's framing of the problem is unusually precise for an agency announcement: "Disordered proteins are like sentences without punctuation — multiple meanings can be implied depending on how the reader pauses or places emphasis."

The first publicly detailed award is NATIVE-ID, funded at approximately $28.6 million. Reading its structure is the closest thing available to a key for the ARPA-H selection process.

The winning team, in full

NATIVE-ID is led by the Innovative Genomics Institute at UC Berkeley — Executive Director Brad Ringeisen, with Hanqin Li heading the Advanced Translational Genetics lab. The named collaborating institutions and investigators:

Eight institutions. Ten named investigators. One award.

The technical core is a measurement effort at a scale that has no real precedent in the field: Halfmann's lab will generate more than 10 billion aggregation measurements across 50,000 proteins using DAmFRET (Distributed Amphifluoric FRET), a single-cell protein self-assembly assay his team published in 2018. That yeast-derived dataset gets paired with human neuron studies, while Brown, Emory, Johns Hopkins, Texas A&M, and Parallel Squared contribute deep-learning frameworks and structural analysis. The initial disease target is frontotemporal lobar degeneration (FTLD), chosen because it shares genetic architecture with ALS.

Halfmann's lab receives approximately $4.1 million over two years — about 14% of the $28.6 million total.

That last number is the most instructive figure in the whole award. The single most technically distinctive component of the project, the thing nobody else could have proposed, is one-seventh of the budget. The other six-sevenths went to AI modeling, human-cell validation, structural analysis, and translational infrastructure built by seven other institutions.

ARPA-H did not fund a capability. It funded an assembled pipeline that ends in a human cell.

The 33-day problem

Reconstruct the solicitation calendar and the implication is inescapable:

Thirty-three days from posting to the solution summary. Seventy-five days to the full proposal. Nineteen days from Proposers' Day to the solution summary.

You cannot assemble an eight-institution, ten-investigator consortium spanning a private research institute, a non-degree-granting genomics institute, two public flagships, three private research universities, and an independent technology institute in thirty-three days. Subawards alone — eight institutional signatures, eight indirect-rate negotiations, eight research-security reviews — take longer than that at most universities.

The NATIVE-ID team existed, in substance, before the solicitation posted. Not as a funded entity, but as a working scientific relationship with pre-negotiated roles.

This is the operational truth about ARPA-H that its outreach calendar actively obscures. Proposers' Day is marketed as the teaming event. By Proposers' Day you have nineteen days left. If you arrive there looking for partners, you are arriving to watch teams that already formed introduce themselves to each other.

The correct inference is not "ARPA-H is closed." It is that ARPA-H's effective teaming window opens when a program is rumored, not when it posts — and that for an agency releasing programs on a rolling basis, the preparation that matters is maintaining a standing consortium that can be pointed at whichever solicitation lands.

What the two technical areas were actually asking for

BIOGAMI's structure explains why the winning shape was a consortium rather than a strong single lab.

TA1 — Molecular grammar of intrinsically disordered regions. Build computational and experimental models defining how intrinsically disordered regions govern protein folding, aggregation, and biomolecular condensate formation. Teams were required to build high-throughput screening platforms and to contribute to open-source ecosystems.

TA2 — Modulation for detection and control. Translate TA1 findings into therapeutic and diagnostic applications. Proposers had to address one rare and one non-rare disease, including at least one neurodegenerative condition, and demonstrate in vivo efficacy by Phase II.

Read TA2's disease requirement as what it is: a portfolio constraint, not a scientific one. ARPA-H wanted each award to hedge across disease prevalence so that the program's results would generalize beyond a single orphan indication. A single-disease proposal — however elegant — was non-responsive on its face.

And the open-source requirement in TA1 is a quiet but real cost. A team that intends to hold its screening platform as proprietary IP was not eligible to win TA1, regardless of how good the platform was. That eliminates a category of otherwise-strong industrial proposers.

The down-select is half the money

BIOGAMI runs 48 months across two sequential 24-month phases. Phase 1 delivers foundational datasets and models. Phase 2 validates therapeutics and early-detection methods. Progression requires meeting metric-driven technical milestones, with a formal down-selection by ARPA-H between phases.

Halfmann's $4.1 million is explicitly described as covering two years — the Phase 1 term. Which means roughly half of the headline $28.6 million is conditional on a down-select ARPA-H has not yet run.

This is standard Other Transaction architecture, and it changes the arithmetic of whether to pursue an ARPA-H award at all. ARPA-H anticipated multiple awards under OT agreements for BIOGAMI, with the solicitation run as an Innovative Solutions Opening (ISO) under full and open competition. OT agreements are what let a non-degree-granting institute prime a $28.6 million biomedical consortium and let an independent technology institute hold a named role. They are also what let the agency stop at month 24 without the appeal structure a grant termination would normally imply.

If your institution's finance office models a 48-month award as 48 months of committed funding, the model is wrong. Model it as a 24-month award with a 24-month option, and staff Phase 1 with people whose positions do not evaporate if the option is not exercised.

Where this applies right now

BIOGAMI is closed. The structural lessons transfer directly to the ARPA-H solicitations that are not.

SURPASS (ARPA-H-SOL-26-164) carries a November 30, 2026 solution summary deadline and a two-stage, 24-to-36-month structure built around a master protocol with an IND gate — including a requirement for multiple intervention arms, for-profit sponsors, and multiple clinical sites. That is BIOGAMI's consortium requirement expressed in clinical-trial form. The same conclusion follows: if you are not already in conversation with the sponsors and sites you would name, November 30 is not enough runway.

SPECTRA (ARPA-H-SOL-26-163) runs staggered solution summary dates across four technical areas, which is a materially friendlier structure — the later technical areas give real teaming time to applicants who start now.

Emerging Health Innovators is the exception that proves the rule: it is explicitly scoped for individual early-career investigators against sixteen named topics, with a terminal-degree cutoff. If you cannot bring eight institutions, EHI is the door that was built for you.

The uncomfortable summary

Three things are true about BIOGAMI simultaneously.

The science ARPA-H bought is genuinely ambitious — 10 billion measurements across 50,000 proteins is a dataset that does not exist anywhere, and building it is the kind of bet a traditional R01 structurally cannot make.

The proposal that won it was a feat of pre-existing coordination at least as much as a feat of science. Eight institutions, ten investigators, thirty-three days.

And half the money is an option, not a commitment.

None of that argues against pursuing ARPA-H. It argues against pursuing ARPA-H reactively. The agency's rolling ISO model rewards organizations that maintain a standing, cross-institutional team with a named lead, pre-cleared subaward terms, and a one-page capability statement ready to reshape — and it quietly penalizes everyone who waits for a posting to start making phone calls.

Pick the two or three ARPA-H mission areas where your work could plausibly land. Find your seven partners now. The next solicitation will give you thirty-three days.