Tips & Tricks for a successful HORIZON-CL6-2027-02-FARM2FORK-08 proposal
Opening
20 April 2027
Deadline
Keywords
Foodomics
AI annotation
Mass spectrometry
Food exposome
Food safety
Machine learning
Farm-to-fork
foodome characterization
Your microfluidic SME partner for Horizon Europe
We take care of microfluidic engineering, work on valorization and optimize the proposal with you
HORIZON-CL6-2027-02-FARM2FORK-08: AI-powered foodome characterization
Food is full of chemistry and we do not know what to do with it. Literally. The number of compounds available in the foods we consume exceeds 100,000, and existing databases monitor only around 150 of them. The Commission would like AI to bridge that divide. It is a matter firmly situated at the intersection of high-throughput analytical chemistry, machine learning, and food safety policy, and it is not anticipated to produce a research report but rather a working tool.
Download the MIC Horizon Europe 2026/2027 Calls Calendar:
Discover more!
Administrative facts: what do we know about the HORIZON-CL6-2027-02-FARM2FORK-08 call?
Which call is it, and when is the opening and the deadline?
- Call name: Call 02 – single stage (2027)
- Call identifier: HORIZON-CL6-2027-02
- Destination: Fair, healthy, and environmentally friendly food systems from primary production to consumption
- Topic: HORIZON-CL6-2027-02-FARM2FORK-08 – AI-powered foodome characterization
- Opening date: 20 April 2027
- Deadline: 23 September 2027
- Type of action: Innovation Action (IA)
What about the budget and estimated size of the project?
- Overall budget: EUR 7.80 million
- Number of projects expected: 1
What are the key eligibility and evaluation conditions?
- Standard eligibility thresholds apply per General Annex B
- Specific restriction: proposals are subject to restrictions for the protection of European communication networks (this one catches people off guard; check the exact scope of that restriction before submitting)
- TRL range: starting from TRL 5, reaching TRL 7 by end of project
- Multi-actor approach: encouraged, not mandatory as an additional eligibility criterion
- JRC participation: not mentioned for this topic
- Clustering requirements: not specified; international cooperation is encouraged
Scientific range: what does the Commission expect from the HORIZON-CL6-2027-02-FARM2FORK-08 grant?
What outcomes are expected?
At the culmination of the project, it should have an approved AI tool that can handle complex foodomics data and recognize and measure chemicals in food far beyond what conventional databases can. It is also the Commission’s desire that there be improved capacity in academia, research infrastructure, and industry to carry out foodomics research, rather than in parallel silos.
What is within scope?
The piece of work has painted a very specific boundary: it is not nutrition policy or diet research, but rather the development of an analytical feature that is powered by AI.
High-throughput foodomics data analysis with AI: untargeted mass spectrometry, nuclear magnetic resonance, and others.
- Creation of AI algorithms that anticipate or conclusively foresee the presence and concentration of compounds, even with incomplete data.
- The phylogenetic relationship between species is applied to enhance inference of compounds.
- Identification of possible food safety hazards by structural similarity to known compounds.
- European research and innovation network building in food exposome research.
- Results of AI-generated foodomics standardization and dissemination.
- Collaboration with the already existing initiatives, e.g., the Centre of Excellence in the Periodic Table of Food Initiative (Wageningen University).
Which are the specifically proposed directions of research?
This is quite explicit as the work program is concerned. The commission is indicating to four angles:
- AI annotation of untargeted data of mass-spec and NMR in analysis with appropriate validation.
- Inference of predictive compounds in the absence of or partial experimental data – this is the technically difficult aspect, and where evaluators are likely to scrutinize most.
- Checking of food safety by similarity of compounds with known dangerous products.
- Network-building between the EU Member States and the Associated Countries to render the resultant data and tools available and interoperable.
Scientific strategy: how can you enhance your chances of being funded through HORIZON-CL6-2027-02-FARM2FORK-08?
What scientific choices matter most?
- Introduce the AI tool as the main character of your proposal. A database or a methodological framework will not satisfy evaluators. At the end, you need a deployable AI demonstrated on actual foodomics datasets.
- Early and hard validation. This is an IA starting at TRL 5. Evaluators will expect a credible validation plan, not an algorithm development roadmap.
- Address the missing-data problem directly. The work program explicitly refers to predicting compound presence from incomplete foodomics data. That is technically non-trivial. Any proposal that sidesteps it will sound weaker than one that names it as a core challenge and proposes a specific plan.
- Deal with the missing-data problem directly. The work program directly refers to the prediction of the presence of compounds based on the unfinished foodomics data. That is technically non-trivial. Any proposal that avoids it will sound weaker than one that mentions it as a fundamental issue and suggests a specific plan of action.
- Relate to the current food safety systems. Structural similarity to known compounds represents the identification of hazards by connecting your AI tool to regulatory relevance. Frame it that way.
- The phylogenetic angle needs to be developed. We would argue this is one of the more novel elements of the call. Any proposal that ignores it misses a real opportunity to stand out.
- Do not treat network activities as an afterthought. The Commission wants a European research network, not just a consortium. Think about how your dissemination and standardization actions will reach beyond your own partners.
Consortium & proposal-writing plan: what works best with this type of call?
- Target somewhere between eight and a dozen partners, maybe a couple fewer if you have strong vertically integrated players. With only one project funded, a tighter consortium is generally preferable to a wide one.
- You should have analytical chemistry laboratories, fitted with mass spec and NMR facilities. No way around it. Those are the generators of data.
- Combine them with AI and bioinformatics teams already active in omics data processing. Food safety or nutrition science expertise alone will not be enough here.
- Have at least one industry partner preferably a food company or an analytical services company. A multi-actor approach is also promoted and the evaluators will observe the absence of industry.
- A new SME focused on AI-based data analysis or food analysis would be a good fit. This is a good position to a smaller and nimble company that has some credibility in this field.
- The international cooperation is promoted explicitly. A collaboration with a North American or APAC organization operating in the food exposome space strengthens your network-building narrative.
- On the proposal itself: start with the gap, not the technology. Food chemistry is mostly unknown territory, and that is a blind spot in population health. Start there, then introduce your AI platform as the answer.
How would microfluidics contribute to this topic?
The scale this topic requires is not the one you build with standard analytical pipelines. Conventional benchtop processes take too long, cost too much, and are hard to standardize when you need to run across hundreds of thousands of compounds and thousands of food samples. Microfluidics combined with AI changes that equation.
- Sample preparation is where foodomics projects lose time. Microfluidic extraction and concentration platforms cut solvent use, reduce processing time, and, most importantly, produce consistent inputs for downstream AI analysis. Garbage in, garbage out, as they say.
- Say you need to screen the same food matrix under fifty production conditions to build a training dataset for your AI algorithm. A droplet-based microfluidic system lets you run those fifty conditions in parallel, same volume, same temperature, same contact time. Same matrix, different compound, same result every time.
- Coupling microfluidic sample handling to mass spectrometry is a maturing technology, and that integration would be directly useful to your consortium, especially for large-scale untargeted analyses. The LOCAI project showed what AI-driven microfluidic automation looks like in practice.
- Manual fractionation steps can also be handled on lab-on-chip platforms. That reduces inter-lab variation, which matters greatly when you are building a network of laboratories across multiple Member States.
- And on that point about networks: the standardization of microfluidic workflows is precisely what makes data interoperable across sites. Your proposal is stronger if it can show not just a smart algorithm, but a pipeline that consistently feeds that algorithm clean, comparable data.
For a call where the Commission wants a validated, deployable AI tool, not a proof of concept, the quality and uniformity of input data is foundational. Microfluidics addresses that directly. Your consortium would be telling a more complete story: not just intelligent analysis, but a controlled, reproducible path from food sample to foodomics dataset. That is the kind of end-to-end pipeline evaluators find convincing.
The MIC already brings its expertise in microfluidics to Horizon Europe:
H2020-NMBP-TR-IND-2020
Microfluidic platform to study the interaction of cancer cells with lymphatic tissue
H2020-LC-GD-2020-3
Toxicology assessment of pharmaceutical products on a placenta-on-chip model
About the author
Julia Sepulveda Diaz, PhD
Chief Executive Officer, Microfluidics Innovation Center
Julia holds a PhD in glycobiology and neurobiology from Université Paris-Est Créteil. Before moving into microfluidics, she spent several years as a postdoctoral researcher at the Paris Brain Institute, on Parkinson’s disease, neuroinflammation, and the role of heparan sulfate in protein aggregation. Today she leads the Microfluidics Innovation Center as CEO, steering its researchers and engineers and its collaborative research and innovation projects.
FAQ – HORIZON-CL6-2027-02-FARM2FORK-08
What makes HORIZON-CL6-2027-02-FARM2FORK-08 different from standard food safety research calls?
This topic is not about detecting known contaminants or validating existing databases. The Commission wants an AI tool that can characterize unknown or poorly documented compounds in food — working with incomplete data and predicting what is likely present based on phylogenetic relationships and structural similarity. That is a genuinely different scientific challenge from most food safety calls.
Is the multi-actor approach mandatory for this call?
No, it is encouraged but not listed as a mandatory additional eligibility criterion. That said, evaluators will almost certainly look for it in practice. A consortium with no industry partner will need a very strong justification to compensate for that gap.
What TRL level does the project need to reach by the end?
The project starts at TRL 5 and must reach TRL 7 by the end. This is an Innovation Action, which means evaluators expect a credible path to deployment — not just a functional prototype in a controlled lab environment.
Can a consortium include non-EU partners in this call?
International cooperation is explicitly encouraged in this topic. Partners from third countries can participate, though they are typically not eligible for EU funding unless specific bilateral agreements apply. Their role is most useful for strengthening the network-building and food exposome coverage narrative — particularly from regions with rich foodomics data.
What AI methods are best suited for foodomics data annotation?
The work programme points to methods suited for untargeted mass spectrometry and NMR data annotation. Graph-based models and transformer architectures have shown strong results in molecular prediction tasks. The ability to handle missing or partial experimental data is the real differentiator — so models with uncertainty quantification built in will score better with evaluators than black-box classifiers.
How important is the phylogenetic angle in this call?
More important than it might look at first glance. Using phylogenetic relationships to infer compound presence across species is one of the more novel directions in the call, and relatively few research teams are actively working on it in a food science context. A proposal that develops this angle credibly will stand out. One that ignores it is leaving points on the table.
What does the Commission mean by food exposome research?
The food exposome refers to the totality of food-related chemical exposures over a person’s lifetime — not just nutrients, but the full chemical landscape of everything ingested. The Commission sees this as a major gap in current public health science. This call is one building block toward closing that gap, specifically by improving the tools available to map and annotate that chemical landscape.
How should we handle the missing-data compound prediction requirement?
Head-on. The work programme names it explicitly, and evaluators will expect a specific technical plan. Approaches that combine structural similarity scoring, phylogenetic inference, and probabilistic modelling tend to perform best in this context. Frame this as a core workpackage, not a secondary feature. It is probably the hardest part of the proposal to make convincing — which also means it is the part that differentiates strong proposals from weak ones.
What existing initiatives should we align our proposal with?
The work programme specifically mentions the Centre of Excellence in the Periodic Table of Food Initiative at Wageningen University. Alignment with that initiative — and potentially with FoodDataCentral (USDA) or similar national food composition databases — strengthens your interoperability story. The Commission wants the outputs to connect with existing infrastructure, not create yet another isolated database.
What are the main evaluation criteria for this Innovation Action?
Standard Horizon Europe evaluation criteria apply: excellence, impact, and quality and efficiency of implementation. For this topic, excellence will focus heavily on the AI methodology and the novelty of the compound prediction approach. Impact will hinge on the network-building dimension and the path to open, interoperable tools. Implementation will be judged on TRL progression credibility and the realism of the validation plan.