HORIZON-HLTH-2027-02-TOOL-01-two-stage: Dates & Page Limits

By Aoba Works · Published October 22, 2026 · 16 min read

How this guide was made: drafted with AI assistance from the official sources linked below, then passed an automated editorial check before publishing. Our process

• Horizon Europe
• Funding And Tenders Portal
• Ria

HORIZON-HLTH-2027-02-TOOL-01-two-stage is a Horizon Europe Health call for AI methods that predict how chronic non-communicable diseases progress and how patients respond to treatment. It opens on 29 October 2026. The Stage 1 deadline is 17 February 2027 and the Stage 2 deadline is 16 September 2027, both at 17:00 Brussels time (Accelopment summary of the official Work Programme; Institut Pasteur). To apply, you submit a 10-page anonymised outline that is scored on Excellence and Impact. If you pass, you are invited to submit a full proposal that is also scored on Implementation. The topic has a budget of EUR 44.2 million, enough for about six projects of EUR 6–8 million each (portal topic data).

Which option is right for you?

The best first step depends on your role. A clinical or academic group should open the topic page, a company should look at how it fits the consortium, and a newcomer should register first.

Coordinating is a heavy job: you manage the Stage 1 submission and then the full proposal over about seven months. If you have never done it before, being a partner is the safer first role.

What is HORIZON-HLTH-2027-02-TOOL-01 about?

The topic funds work that uses established AI methods to find, validate and deploy predictive biomarkers for chronic non-communicable diseases. Cancer is excluded (portal topic data). The full title is "Development of predictive biomarkers of disease progression and treatment response by using AI methodologies for chronic non-communicable diseases". It is a Research and Innovation Action (HORIZON-RIA).

The official text lists four expected outcomes. Paraphrased:

1. Clinical researchers get access to novel predictive biomarkers for assessing disease progression and treatment response more accurately.
2. Clinicians use clinically validated predictive biomarkers in clinical research and personalised medicine.
3. Stakeholders get trustworthy AI tools that guide the development of multimodal predictive biomarkers.
4. Citizens benefit from better clinical guidelines and biomarker-guided clinical research.

The scope then asks for five activities. Reviewers will look for each of them:

• Map what exists. Set up a multidisciplinary collaboration that maps and evaluates the biomarker data already used in clinics. Include stratification by biological sex and, where feasible, gender-related and sociodemographic variables.
• Adapt, don't invent. The text says to "adapt and apply established AI methods rather than developing novel ones from scratch". These methods should run on multimodal biomarkers (molecular, cellular, physiological, imaging, behavioural, digital) combined with longitudinal clinical studies and real-world data (RWD).
• Model relationships with trustworthy AI. Describe how biomarkers relate to each other, with a focus on AI trustworthiness, plus "user-friendly and fit-for-purpose visualisation and decision-support tools".
• Validate clinically. Build a biomarker validation platform and run validation studies in independent disease cohorts, RWD and exploratory clinical studies.
• Plan exploitation and regulation. Write an exploitation plan and a regulatory strategy, with priority on exploiting results in the EU.

On top of these five, projects must follow GDPR-compliant data protection practice. They are encouraged to use research and data infrastructures funded under the Digital Europe Programme. They should also set aside budget for networking and joint activities with other funded projects (portal topic data).

Example (illustrative only): a consortium working on chronic kidney disease could pool longitudinal registry data, proteomics and wearable data from three national cohorts. It would fine-tune existing gradient-boosting and survival models, then validate the predicted progression risk in a fourth cohort it did not train on. The tool would be a clinician dashboard. A proposal like this matches all five activities. A proposal built around a new foundation model architecture does not match the second one.

What is the deadline for Stage 1 and Stage 2?

Stage 1 closes on 17 February 2027 and Stage 2 on 16 September 2027, both at 17:00 Brussels local time. The call opens on 29 October 2026 (Accelopment, reporting the Work Programme published on 1 October 2026).

These dates are newer than the draft dates. The draft Work Programme had the call opening on 10 February 2027, with Stage 1 on 13 April 2027 and Stage 2 on 22 September 2027. The final version moved everything earlier (Accelopment on the draft changes). Some third-party pages still show the old April dates, for example this tips page. Plan around 17 February. The Work Programme also lets the responsible Director-General delay deadlines by up to two months (Institut Pasteur summary). Don't build your plan on that possibility.

One trap: as of 22 October 2026 the topic status on the portal is "Forthcoming". The topic data feed we read still showed placeholder dates (June and September 2026) that don't match the Work Programme (portal topic data). Once the call opens, the deadline shown in the Submission service is the one that counts.

What are the page limits for Stage 1 vs Stage 2?

Stage 1 has a 10-page limit. A full Stage 2 RIA proposal has a 40-page limit, or 45 pages if the topic uses lump-sum funding (General Annexes 2026-2027; summarised by criteri.ai). The topic conditions we read don't mention lump sums (portal topic data), so 40 pages is the likely limit. Confirm it in the Stage 2 template, which is generated for the topic.

The 2026–2027 application form (V5.0 from November 2025, V5.1 from January 2026) reduced the page allowance and simplified sections 2.1 and 3.1 (criteri.ai, citing the official version history; Accelopment on template changes). Old Part B files from 2024 calls should only be used for reference.

Common pitfalls and fixes:

• Going over the limit. Evaluators ignore pages beyond the limit. Fix: freeze the outline at 9 pages a week before the deadline, which leaves room for late edits.
• Identifying yourself in Stage 1. First-stage proposals in the 2026–27 programme are evaluated blind. Part A data is hidden automatically, but removing names from Part B and the abstract is your job, and a direct identification makes the proposal inadmissible (summary of General Annexes rules). Fix: write "Partner 1 (university hospital, 4,000-patient CKD cohort)" and never "Charité cohort". Also remove named software, grant numbers and self-citations that reveal who you are.
• Using the wrong template. Fix: download Part B from inside your draft in the Submission service, not from an old shared drive.

Who can apply to this call (eligibility)?

Under the standard Horizon Europe RIA rule in the General Annexes, you need at least three independent legal entities from three different EU Member States or Associated Countries, with at least one in a Member State. The topic adds one rule of its own: prospective clinical studies must be led by entities in the EU/EFTA and/or Associated Countries (portal topic data; General Annexes). The topic refers to Annex B of the General Annexes for eligible countries and Annex C for financial and operational capacity.

US entities can receive funding here. The Health cluster opens Union funding to US-based legal entities in recognition of NIH funding European researchers (Science|Business on EU–NIH reciprocity). Check the "Conditions" section on the topic page once the call opens to confirm this still applies. Partners from other non-associated countries usually join at their own cost.

Steps to check eligibility before you write anything:

1. List every partner with its country and legal status (university, hospital, SME, NGO, public body).
2. Confirm that at least three are independent of each other and based in three different EU/Associated countries.
3. Mark which partner leads each prospective clinical study and confirm it is EU/EFTA/Associated-based.
4. Make sure every partner has a 9-digit PIC. If not, follow the PIC registration guide.
5. Check whether your coordinator is likely to pass the financial capacity check (Annex C). Coordinators of grants of EUR 6–8 million get the most scrutiny.

Pitfall: two hospital departments of the same university group count as one entity if they are not independent. Fix: add a separate entity from a third country.

What is the evaluation criteria for Stage 1?

At Stage 1, experts score only Excellence and Impact, and each criterion needs at least 4 out of 5 (portal topic data). That is stricter than the usual full-proposal threshold of 3, and the bar can rise further.

The General Annexes also set an overall threshold on the sum of the two scores. It is chosen so that the total budget requested by proposals going to Stage 2 is close to three times the available budget, and never less than twice. In practice it is "expected normally to be set at 8 or 8.5" (General Annexes rules, as summarised).

With EUR 44.2 million available, that funnel means roughly EUR 130 million of proposals reach Stage 2, or about 16–22 consortia at EUR 6–8 million each (example arithmetic, not an official figure). For about six grants, that works out to around one in three invited proposals being funded.

What scores 4+ at Stage 1 for this topic, based on the scope:

• A named disease area (or a clearly argued cluster of NCDs) and a clear gap in current predictive biomarkers.
• Evidence you have the data: cohort sizes, follow-up length, data types, and an independent validation cohort. Describe these without naming the institutions.
• An explicit statement that you adapt established AI methods, with the trustworthiness and bias work you will do.
• A credible route to clinical use: guidelines, regulatory strategy, and exploitation in the EU.

For section-by-section scoring prompts, use the Part B scoring map.

What is the evaluation criteria for Stage 2?

At Stage 2, experts score Excellence, Impact and Implementation. Each needs at least 4 out of 5, and the total must reach 12 out of 15 (portal topic data). Generic guides often quote the standard 3/5 per criterion and 10/15 overall. Those numbers do not apply to this topic.

Stage 2 must build on what you promised at Stage 1. If you change the objectives a lot, reviewers will notice a mismatch with the outline they invited. You can strengthen the consortium, for example by adding an SME for the decision-support tool, as long as the core concept stays the same.

Pitfalls and fixes at Stage 2:

• Treating the clinical studies annex as optional. The topic requires it (portal). Fix: draft it in parallel with the work plan, starting in spring 2027.
• No budget for networking. The scope asks for "sufficient budget for networking and joint activities". Fix: put a specific person-month and travel line in a work package.
• A vague regulatory plan. Fix: name the regulatory route your tool would likely follow and when you plan to engage regulators.

How do I submit a proposal via the Funding & Tenders Portal?

You submit from the topic page by clicking "Start submission". This opens the Submission wizard, where you create the draft, add partners, complete Part A and upload Part B before 17:00 Brussels time (Online Manual: Electronic proposal submission; IT How To: Submit a proposal).

1. Create an EU Login account for every person who will edit the proposal.
2. Get a PIC for each organisation in the Participant Register (step-by-step guide).
3. Find the topic. On the Funding & Tenders Portal, go to "Search Funding & Tenders" and search for HORIZON-HLTH-2027-02-TOOL-01-two-stage. The "Start submission" button appears once the status changes from "Forthcoming" to "Open" (expected 29 October 2026).
4. Click "Start submission" and choose the action type HORIZON-RIA when asked.
5. Pre-register the draft. The coordinator enters the acronym and short summary. Keep the summary anonymous, because it is also covered by blind evaluation.
6. Add participants. In the participants step, add each partner by PIC and give contact persons their access roles.
7. Complete Part A (the administrative forms): budget table, declarations, ethics and security self-assessment.
8. Download the Part B template from inside the draft, write it, export it to PDF and upload it.
9. Click "Validate" and fix every error the system flags (missing fields, page count, file format).
10. Click "Submit" and check that you receive the acknowledgement of receipt. You can resubmit as often as you want before the deadline. Only the last version is evaluated.

For Stage 2, only invited coordinators can open the second-stage draft. Partners and Part A data are carried over and you can edit them.

Common mistakes:

• Submitting at 16:55. The portal is busy on deadline day. Fix: submit a complete version 48 hours early, then overwrite it if needed.
• Partners without PICs on deadline week. Fix: collect PICs in November 2026, well before the deadline.
• Identifying metadata in the PDF. File properties can contain author names. Fix: clear the document properties before exporting.

What is the indicative budget and project duration for this call?

The topic budget is EUR 44.20 million for 2027, with an expected EU contribution of EUR 6–8 million per project and about six grants (portal topic data; Accelopment). The topic text we read sets no fixed duration, so you choose one and justify it in the work plan.

Because this is an RIA, the EU normally reimburses 100% of eligible costs, plus a 25% flat rate for indirect costs (General Annexes). You can ask for more than EUR 8 million, but "the range" signals what evaluators expect. A request far outside it has to be argued clearly.

Example budget split (illustrative, not official guidance): for a EUR 7 million request over 48 months, a consortium might put about 40% into clinical validation (cohort access, exploratory studies), 25% into AI modelling and trustworthiness, 15% into the decision-support tool and its usability, 10% into exploitation and regulation, and 10% into management, dissemination and networking. Validation-heavy splits like this match the scope's weight on clinical utility.

AI-driven biomarkers vs traditional biomarkers: what does this call expect?

The call wants established AI/ML methods applied to multimodal data and validated clinically. A single-marker study using classic statistics doesn't fit, and neither does inventing a new AI architecture (portal topic data). The difference is real, and many proposals get it wrong in one direction or the other.

Example of fit (illustrative): a Parkinson's disease project that combines digital gait data from smartphones, MRI and CSF markers, uses existing deep survival models, and checks performance separately for women and men. This matches the topic. A project whose main deliverable is a new transformer architecture for omics belongs elsewhere. The 2027 Health programme has a separate topic, HORIZON-HLTH-2027-03-TOOL-08, on frontier AI models for healthcare (Accelopment).

Pitfalls:

• "AI-washing". Mentioning machine learning without saying which models, which data and which validation metric. Fix: name the model family, the training and validation cohorts, and the clinical endpoint.
• Cancer creeping in. Cancer is out of scope. Fix: keep oncology comorbidities out of your main endpoints.
• No plan for clinicians to use it. Fix: put usability testing of the decision-support tool with clinicians into a work package.

Next steps

You can start today, three and a half months before the Stage 1 deadline. Here is what to do yourself, what you could hand off, and where to read next.

Do it yourself today:

1. Check the PIC of every likely partner and ask anyone without one to register this week (guide).
2. Write a one-page anonymised concept that covers the disease area, data modalities, cohort sizes, independent validation cohort and the clinical decision your tool supports. Test it against the five scope activities above.
3. Put 29 October 2026 (call opens), 17 February 2027 and 16 September 2027 in your shared calendar, with an internal Stage 1 submission date of 13 February.

If you'd rather not do it alone:

• National Contact Points (NCPs) for Cluster 1 Health are free, official advisers in each country. They review eligibility and explain the rules (Health NCP network). They suit first-time coordinators. They are not ghost-writers: they won't write your Part B, and how much time they can give varies by country.
• Private grant consultancies can write or edit proposals for a fee. Ask for a written quote that sets out the full fee structure before you commit. They suit consortia that have strong science but little proposal-writing time. They are a poor fit if you can't share confidential data or can't afford fees on a call where most applicants are not funded. A consultant still can't create a validation cohort you don't have.

Related guides on this site:

• Horizon Europe Proposal Template Part B: 2026 Scoring Map
• EU Funding for SMEs: 7 Open 2026 Calls and How to Qualify
• EIC Accelerator Deadlines: 2026 Cut-Off Dates and Steps
• Cascade Funding Open Calls: Portal Filters and 2026 Dates

FAQ

What are common reasons proposals get rejected at Stage 1?

The usual reasons are missing the 4/5 threshold on Excellence or Impact, or landing under an overall cut-off that is normally 8 or 8.5. For this topic, typical weak points are no independent validation cohort, proposing new AI methods instead of adapting established ones, and no sex/gender stratification. Identifying your organisation in Part B or the abstract breaks the blind-evaluation rule and can make the proposal inadmissible.

Can a project resubmit if not invited to Stage 2?

Nothing in the topic text bars you from submitting an improved version to a later suitable call. But this topic appears only in the 2027 call of the 2026–2027 Work Programme, so there is no second cut-off for it. Use the evaluation feedback to rework the idea for another Health topic, and check each new call's own rules.

Does this call require patient data management plans?

The topic requires GDPR-compliant personal data protection and AI and datasets that comply with privacy-preserving law (portal topic data). The standard RIA Part B also asks how you will manage research data (annotated RIA/IA template). At Stage 2 you also complete the clinical studies annex, which covers how patient data are collected and handled.

Is there a specific TRL (Technology Readiness Level) requirement?

No. The topic conditions we read do not set a TRL. Because this is an RIA with clinical validation and exploratory studies, evaluators will expect the project to move biomarkers and tools from proven concept towards clinically tested prototypes. Describe the starting point and the end point clearly in the Excellence section.

Sources

• Funding & Tenders Portal: topic data for HORIZON-HLTH-2027-02-TOOL-01-two-stage
• Horizon Europe Work Programme 2026-2027, 4. Health (European Commission)
• Horizon Europe Work Programme 2026-2027, 15. General Annexes
• Online Manual: Electronic proposal submission
• IT How To: Submit a proposal
• Accelopment: Horizon Europe Health 2027 calls, official Work Programme published
• Accelopment: Draft Health Work Programme 2027 deadline changes
• Accelopment: 2026–2027 application template changes
• Institut Pasteur: Health Cluster February 2027 deadlines
• criteri.ai: Page limits and Part B template 2026-2027
• criteri.ai: Evaluation criteria scoring guide
• Microfluidics Innovation Center: Tips for HORIZON-HLTH-2027-02-TOOL-01 (draft dates)
• Science|Business: NIH reciprocity and EU health funding
• Health NCP Net: Cluster 1 Health
• Annotated RIA/IA template, 2026-27 topics (Horizon Europe NCP Portal)

Spotted something out of date or wrong? Tell us at shigoto.ai.note@gmail.com. Guides are re-checked against official sources on a rolling schedule.