NIH Just Collapsed Its Entire R25 Landscape Into Three Parent NOFOs — and the First Deadline Is September 25

If you run an NIH-funded research education program — a summer undergraduate research experience, a short course in statistical genetics, a mentoring network for early-career clinician-scientists — the announcement you have been submitting to for the last several years is very likely dead. Not expiring on schedule. Expired early, effective immediately, by notice.

In its place are three parent Notices of Funding Opportunity, released August 25, 2026, that now carry essentially all of NIH's R25 research education portfolio:

• PA-27-034 — NIH Research Education Program: Courses, Curriculum & Methods (Parent R25 — Clinical Trial Not Allowed)
• PA-27-035 — NIH Research Educational Program: Mentoring Activities & Networks (Parent R25/UE5 — Clinical Trial Not Allowed)
• PA-27-036 — NIH Research Education Program: Research Experiences (Parent R25 — Clinical Trial Not Allowed)

All three expire May 26, 2029. All three run on the same three-cycle calendar: September 25, January 25, May 25, annually, with applications due by 5:00 PM local time of the applicant organization.

Which means the first submission window under the new regime closes September 25, 2026 — two days from this writing.

This is the same structural move NIH made to the K-award family earlier this year, when it announced the consolidation of career development mechanisms into four codes under NOT-OD-26-098 (our analysis of that change). The R25 consolidation is faster, quieter, and — because the institute-specific announcements are being killed rather than sunset — considerably more disruptive to programs mid-cycle.

What actually died

The consolidation is being executed through a wave of early-expiration notices, each one short and each one removing a NOFO that a specific research community had organized itself around:

• NOT-CA-26-017 expired NCI's Cancer Research Education Program announcements PAR-23-276, PAR-23-277 and PAR-23-278, effective immediately, directing new, renewal and resubmission applications to the three new parents.
• NOT-DA-26-015 expired PAR-25-204, NIDA's Research Education Program for Clinical Researchers and Clinicians.
• NOT-GM-26-010 expired PAR-24-252, NIGMS's Innovative Programs to Enhance Research Training (IPERT).

Note what those three have in common. They were not generic. PAR-23-276/277/278 were the cancer education pipeline. PAR-25-204 was specifically aimed at clinicians moving into addiction research. IPERT was NIGMS's flagship vehicle for research training innovation. Each carried its own review panel culture, its own institute-specific evaluation expectations, and — critically — its own program officers who knew the applicant pool.

Consolidation into a parent NOFO does not preserve any of that automatically. It replaces a targeted announcement reviewed against a targeted portfolio with a general announcement reviewed against everything.

The three buckets, and why the sorting is the hard part

The new architecture divides research education by activity type, not by disease area or institute. That is a genuinely different taxonomy than most program directors have been working in, and it forces a decision before you write a word.

PA-27-034 (Courses, Curriculum & Methods) covers two named components: Courses for Skills Development, and Curriculum or Methods Development. This is where a short course in single-cell analysis goes. It is where a new graduate methods sequence goes. It is where a modular curriculum you intend other institutions to adopt goes.

PA-27-036 (Research Experiences) is the hands-on bucket — summer programs, structured research placements, the whole category of "put a learner on a bench and supervise them." NIH lists 18 participating Institutes and Centers on this one, the broadest of the three.

PA-27-035 (Mentoring Activities & Networks) is the newest conceptual category and the one with the most distinctive machinery. Its primary component is Mentoring Activities: programs that strengthen professional development through mentoring, and programs that improve the quality of mentorship itself. Sixteen ICs participate.

The sorting problem is real because most mature research education programs do all three. A summer research program with a methods bootcamp and a mentor-training module maps onto all of 034, 035 and 036. The new structure asks you to declare a primary focus and build the rest as supporting activity. Get that wrong and you land in a study section evaluating your application against a different center of gravity than the one your program actually has.

PA-27-035 has a mechanism the others do not

Only PA-27-035 is a Parent R25/UE5. Participating ICs may elect to make an award under that announcement as a UE5 cooperative agreement rather than an R25 grant — an assistance mechanism in which substantial federal scientific or programmatic involvement is anticipated throughout the performance period.

The NOFO is careful about the boundary: NIH staff serve in an advisory capacity and do not direct activities, and the recipient retains what the announcement calls "prime responsibility" for implementation. But a cooperative agreement is still a materially different operating posture than a grant. It means standing program calls, negotiated milestones, and an NIH project scientist with a defined role in your governance.

If you are building a multi-site mentoring network — the exact use case PA-27-035 seems designed to absorb — you should assume the cooperative agreement path is live and staff your budget for the coordination overhead it implies. If you want a straight grant with minimal federal involvement, that is a conversation to have with a program officer before submission, not after.

The 8% indirect cost ceiling is the number that will surprise people

Both PA-27-035 and PA-27-036 state the R25 facilities and administrative cost allowance explicitly: 8% of modified total direct costs, exclusive of tuition and fees, expenditures for equipment, and consortium costs in excess of $25,000.

Eight percent. Not your negotiated rate. For an institution with a 58% federally negotiated rate, an R25 recovers roughly one-seventh of what an R01 of the same direct-cost size would recover.

This is not new to R25s — the education-grant F&A limitation has been longstanding NIH policy — but it is newly visible now that every R25 in the system routes through three announcements that say it in plain text. And it is arriving in a year when indirect cost recovery has been the single most contested line item in federal research funding.

The practical consequence: research education programs are, and will remain, institutionally subsidized. When your dean evaluates whether to support a resubmission, the honest arithmetic is that an R25 at the expected ceiling of $500,000 direct costs over a project period of up to five years brings in roughly $40,000 in recovery across the life of the award. The case for doing it is mission, pipeline and reputation. It is not overhead.

Budget guidance across the three announcements is consistent: application budgets are not formally capped but "must reflect the actual needs of the proposed project," with the expectation that they do not exceed $500,000 in direct costs. Project periods run to a maximum of five years. No cost sharing is required.

Two eligibility rules that quietly narrow the pool

Participants should be primarily US citizens and permanent residents. PA-27-036 states it directly: research education programs should be used primarily for the education of US citizens and permanent residents. For institutions whose graduate and postdoctoral populations skew heavily international, this is a design constraint on recruitment, not a footnote — and one your evaluation plan will be read against.

Foreign organizations are explicitly ineligible. Non-domestic entities cannot apply. Domestic applicants may still propose consortium arrangements, but the prime must be US-based.

Eligible organizations are otherwise broad: higher education institutions of all types, nonprofits, for-profits, state and local governments, tribal governments, faith-based organizations, and independent school districts. NIH also explicitly encourages applications from institutions in IDeA-eligible states, HBCUs and Tribal Colleges and Universities.

The PD/PI bar is looser than many applicants assume — any individual with the skills, knowledge and resources to carry out the proposed work — with one substantive condition: at least one PD/PI must have a record of using rigorous biomedical research methods. All PD/PIs must register with ORCID and maintain active eRA Commons accounts.

The complementarity rule is the most common reason these applications fail

Every one of the three announcements carries a version of the same requirement: the proposed activities must be distinct from research training and research education programs currently receiving federal support at the applicant institution.

This is the requirement that kills otherwise strong applications. If your institution holds a T32 in the same scientific area, a reviewer will ask — and is instructed to ask — what the R25 does that the T32 does not. "More students" is not an answer. "Different population, different stage, different skill, different institutional partner" is.

Write that comparison into the application explicitly. Name the existing federal awards. Draw the boundary yourself, in your own words, before a reviewer draws it less generously.

What has to be in the Research Education Program Plan

The required structure is consistent and detailed. Expect to produce:

• A description of the proposed research education program
• PD/PI qualifications and program faculty credentials
• Participant eligibility criteria and a recruitment strategy
• Documentation of institutional environment and commitment
• A Plan for Instruction in Methods to Enhance Reproducibility
• A Plan for Instruction in the Responsible Conduct of Research
• An evaluation plan with baseline metrics
• A dissemination plan for materials and findings

Two of those deserve emphasis. The reproducibility instruction plan is a named, separately reviewed component — not a paragraph folded into the curriculum description. And the evaluation plan requires baseline metrics, meaning you need to state what you are measuring against before the program starts, not a list of outputs you intend to count afterward.

Scored review criteria are the standard five: Significance, Investigator(s), Innovation, Approach, Environment. Applications are evaluated on the likelihood of strongly advancing research education through the stated program goals and participant outcomes.

There is no clinical-trial-allowed companion

All three parents are Clinical Trial Not Allowed. The announcements do permit human-subjects activities that do not meet the NIH clinical trial definition — an important distinction, since a great deal of education research involving learners is human-subjects work without being a trial.

But if your research education program embeds an independent clinical trial as a training vehicle, there is currently no parent R25 that will take it. That is the same directional move NIH made in the K-award consolidation, where the independent-clinical-trial variants were eliminated rather than preserved. Treat it as policy, not oversight, and plan the trial component under a separate mechanism.

What to do in the next seventy-two hours

If you were planning a September 25 submission to an institute-specific R25 that has now expired, you have three realistic options.

Submit to the parent this cycle. The due date is identical to the one you were already building toward. The substantive rewrite is mostly structural: pick your primary bucket, restate the complementarity argument against currently funded federal programs, confirm the budget clears at 8% F&A, and verify the participant composition language.

Slip to January 25, 2027. Four months is enough time to do the bucket-selection question properly rather than under duress, and the parent announcements run through May 2029 — there is no scarcity argument for rushing.

Talk to a program officer first. This matters more than usual right now. The ICs that just expired their own announcements have institutional views about which parent bucket their former applicants should land in, and those views are not written down anywhere. A fifteen-minute call is the highest-return activity available to you this week.

NIH has scheduled a webinar on November 5, 2026 at 2:00 PM ET to walk through the three new opportunities. That is useful for the January and May cycles. It is three weeks too late for anyone deciding what to do about September 25 — which is, in a sense, the whole story of how this consolidation has been rolled out.