Ettevalmistustoetuse taotlemine - BioMedGen

5873115 428625 0 0 -3810 9724390 TALLINNA TEHNIKAÜLIKOOL Ehitajate tee 5 19086 Tallinn Rg-kood 74000323 Tel 620 2002 E-post [email protected] www.taltech.ee 0 0 TALLINNA TEHNIKAÜLIKOOL Ehitajate tee 5 19086 Tallinn Rg-kood 74000323 Tel 620 2002 E-post [email protected] www.taltech.ee -3810 428625 0 0 Sihtasutus Eesti Teadusagentuur Soola 8 0 3 . 0 3 .202 6 nr 11-10/48- 4 51004 Tartu Ettevalmistustoetuse taotlemine Käesolevaga taotleb Tallinna Tehnikaülikool ( reg nr 74000323, ak EE201010052037382001 ) ettevalmistustoetust projektile Empowering BIOMEDical Science with Multimodal GENerative AI Technologies – BioMedGen ( 101290382 , HORIZON-HLTH-2025-01 ), vastutav täitja Pirjo Spuul, keemia ja biotehnoloogia instituut. Lugupidamisega (allkirjastatud digitaalselt) Marika Lunden Teadusosakond Lisa: hindamisleht left 284480 Riina Vilgats 620 3536 [email protected] 0 0 Riina Vilgats 620 3536 [email protected] Associated with document Ref. Ares(2026)1802897 - 17/02/2026 Proposal Evaluation Form EUROPEAN COMMISSION Evaluation Summary Horizon Europe (HORIZON) Report - Research and innovation actions Call: HORIZON-HLTH-2025-01 Type of action: HORIZON-RIA Proposal number: 101290382 Proposal acronym: BioMedGen Duration (months): 60 Proposal title: Empowering BIOMEDical Science with Multimodal GENerative AI Technologies Activity: HORIZON-HLTH-2025-01-TOOL-03 N. Proposer name Country Total % Grant % eligible Requested costs 1 TALLINNA TEHNIKAÜLIKOOL EE 1,452,313.75 8.59% 1,452,313.75 8.59% 2 LUDWIG-MAXIMILIANS-UNIVERSITAET MUENCHEN DE 0 0.00% 0 0.00% 3 KLINIKUM DER LUDWIG-MAXIMILIANS-UNIVERSITAT DE 1,260,915 7.45% 1,260,915 7.45% MUNCHEN 4 MAX-PLANCK-GESELLSCHAFT ZUR FORDERUNG DER DE 1,274,111.25 7.53% 1,274,111.25 7.53% WISSENSCHAFTEN EV 5 LINAC-PET SCAN OPCO LIMITED CY 400,000 2.36% 400,000 2.36% 6 PROTOBIOS OU EE 497,750 2.94% 497,750 2.94% 7 SIB SWISS INSTITUTE OF BIOINFORMATICS CH 915,250 5.41% 915,250 5.41% 8 NACIONALNI INSTITUT ZA BIOLOGIJO SI 1,006,000 5.95% 1,006,000 5.95% 9 NEC LABORATORIES EUROPE GMBH DE 630,640 3.73% 630,640 3.73% 10 NEC ONCOIMMUNITY AS NO 286,250 1.69% 286,250 1.69% 11 CSEM CENTRE SUISSE D'ELECTRONIQUE ET DE CH 758,880 4.49% 758,880 4.49% MICROTECHNIQUE SA - RECHERCHE ET DEVELOPPEMENT 12 LATVIJAS UNIVERSITATE LV 665,750 3.94% 665,750 3.94% 13 FUNDACIO EURECAT ES 862,125 5.10% 862,125 5.10% 14 STICHTING IMEC NEDERLAND NL 709,632.5 4.19% 709,632.5 4.19% 15 FUNDACIO PRIVADA PER A LA RECERCA I LA ES 78,000 0.46% 78,000 0.46% DOCENCIA SANT JOAN DE DEU 16 PARC SANITARI SANT JOAN DE DEU ES 392,625 2.32% 392,625 2.32% 17 INSTITUTO PEDRO NUNES ASSOCIACAO PARA A PT 545,625 3.23% 545,625 3.23% INOVACAO E DESENVOLVIMENTO EM CIENCIA E TECNOLOGIA 18 UAB TERAGLOBUS LT 415,000 2.45% 415,000 2.45% 19 UNIVERSIDAD DEL PAIS VASCO/ EUSKAL HERRIKO ES 649,870 3.84% 649,870 3.84% UNIBERTSITATEA 20 I+MED S COOP ES 740,292.5 4.38% 740,292.5 4.38% 21 Nexomic Limited IE 599,937.5 3.55% 599,937.5 3.55% 101290382/BioMedGen-17/02/2026-13:58:49 1/5  Associated with document Ref. Ares(2026)1802897 - 17/02/2026 22 BARCELONA SUPERCOMPUTING CENTER CENTRO ES 500,000 2.96% 500,000 2.96% NACIONAL DE SUPERCOMPUTACION 23 SIA GREMOSANAS SLIMIBU CENTRS GASTRO LV 462,500 2.73% 462,500 2.73% 24 CHINO SRL IT 456,000 2.70% 456,000 2.70% 25 Forsante Oy FI 558,687.5 3.30% 558,687.5 3.30% 26 LIETUVOS SVEIKATOS MOKSLU UNIVERSITETAS LT 557,505 3.30% 557,505 3.30% 27 ENOSI ASTHENON ELLADAS EL 241,000 1.42% 241,000 1.42% Total: 16,916,660 16,916,660 Abstract: Europe faces a major healthcare challenge from complex, chronic, and multifactorial diseases, particularly mental disorders. These conditions affect tens of millions, reduce life expectancy, and impose an economic burden exceeding €500 billion annually. Prevalence among adolescents is rising, while current treatments often fail to deliver personalised, predictive, and continuous care. Despite advances in biomarker research and digital health, key gaps still remain including delayed detection, variable treatment outcomes, and limited system capacity for personalised, predictive care. Critical biological mechanisms, including immune dysregulation, chronic inflammation, metabolic imbalance, and gut–brain–immune interactions, remain only partially understood. Progress is hindered by siloed disciplines and poor interoperability of health data. At the same time, Europe generates petabyte-scale multimodal data from genomics, imaging, wearables, and electronic health records, yet only a fraction is exploited. Conventional AI cannot manage such heterogeneous data and often fails to meet European requirements for transparency, interpretability, and regulatory compliance. BioMedGen will design, develop, and validate Biomedical Generative AI (GenAI) ecosystem services to enable secure, ethical, and scalable GenAI development, deployment, and governance across Europe. Portable and regulation-ready, these services will interoperate with EU research infrastructures. A core component, the GenAI Sandbox, will provide a secure environment for co-designing, testing, and validating models, enabling federated learning, workflow orchestration, and rigorous regulatory-grade evaluation. We will also provide a library of core AI components and six specialized tool suites, first piloted in local research environments and then integrated into the Sandbox. Data governance and privacy-preserving techniques will ensure GDPR and AI Act compliance. Evaluation Summary Report Evaluation Result Total score: 13.50 (Threshold: 12 ) Criterion 1 - Excellence Score: 4.50 (Threshold: 4 / 5.00 , Weight: - ) 101290382/BioMedGen-17/02/2026-13:58:49 2/5  Associated with document Ref. Ares(2026)1802897 - 17/02/2026 The following aspects will be taken into account, to the extent that the proposed work corresponds to the description in the work programme: - Clarity and pertinence of the project’s objectives, and the extent to which the proposed work is ambitious and goes beyond the state of the art. - Soundness of the proposed methodology, including the underlying concepts, models, assumptions, inter-disciplinary approaches, appropriate consideration of the gender dimension in research and innovation content, and the quality of open science practices, including sharing and management of research outputs and engagement of citizens, civil society and end users where appropriate. The project objectives are clear, coherent, and pertinent to the topic on trustworthy generative AI (GenAI) for biomedical and clinical research through their focus on the generation of several generative AI models using different use cases across several medical fields (psychiatry and oncology). The proposal provides measurable indicators, such as area under the receiver operating curve (AUROC) thresholds and calibration metrics. The proposed work is highly ambitious, integrating multimodal GenAI, federated learning, synthetic data, explainability, and connections with major European infrastructures (EOSC, EUCAIM, BBMRI-ERIC). The proposal establishes a clear baseline of the prior state of the art in terms of technical performance. The concept of a unified GenAI Sandbox or multi-agent framework is innovative,with the potential to advance beyond current fragmented AI pipelines in biomedical research. The use cases, including glioblastoma-on- chip (GBMoC) integration and the digital gut twin, represent innovative developments that go beyond current practice. The R&I maturity is appropriate, with the project progressing from conceptual AI modules toward integrated, validated multimodal systems. The expected maturity level (TRL 2-6) is consistent with the research-oriented nature of the proposal. Overall, the methodology is sound and clear. The development of AI tools is well described (model development, federated learning, explainability, bias control). However, some aspects of the methodology related to biomarker validation studies are not adequately addressed. For example, the validation of biomarkers for patient stratification in psychiatric diseases, using only in vitro models rather than clinical samples, is not convincing. This is a shortcoming. Relevant national and international research and innovation initiatives are convincingly integrated into the methodology. Alignment with European infrastructures, standards, and ongoing initiatives is a methodological strength. The proposal brings together a broad range of disciplines including clinicians, biomedical researchers, oncologists, AI/machine learning (ML) developers, data scientists, regulatory specialists, ethicists, sociologists, and patient representatives, and outlines mechanisms for integrating these areas. The proposal adequately integrates social sciences and humanities. The gender dimension is appropriately considered in the research methodology. Sex- and gender-based analyses are planned within the use cases, alongside attention to gender balance in recruitment, cohort composition, data analysis, bias mitigation, and dissemination. Open science and FAIR principles are well reflected in the methodology. The proposal outlines open licensing of results, code, and models, and the use of standards -such as OMOP, FHIR, GA4GH, and RO-Crate- to support interoperability and alignment with EHDS. The planned use of open repositories (e.g., GitHub, Zenodo with DOIs) supports transparency and reuse. Research data management is appropriately described across the different elements of the proposal. Criterion 2 - Impact Score: 4.00 (Threshold: 4 / 5.00 , Weight: - ) The following aspects will be taken into account, to the extent that the proposed work corresponds to the description in the work programme: - Credibility of the pathways to achieve the expected outcomes and impacts specified in the work programme, and the likely scale and significance of the contributions from the project. - Suitability and quality of the measures to maximise expected outcomes and impacts, as set out in the dissemination and exploitation plan, including communication activities. The project provides a generally credible contribution to the expected outcomes and impacts of the topic by developing trustworthy GenAI tools aimed at strengthening biomedical and clinical research capacity in Europe. Pathways to achieving the expected outcomes are articulated through the development of the GenAI Sandbox, validation in four clinical studies, alignment with major EU infrastructures (EOSC, AI Factories, EUCAIM, BBMRI-ERIC), and the use of open calls for training to broaden adoption. The proposed work has strong clinical and societal relevance, particularly in oncology and early cancer detection, and the ethical, regulatory, and governance frameworks are robust. The proposal's results will contribute significantly to some of the expected impacts. The project’s alignment with European digital-health and data infrastructures supports longer-term impact prospects, and the inclusion of training and capacity-building activities enhances potential sustainability. Strong policy alignment supports long-term adoption. However, long-term sustainability mechanisms are unclear. In particular, post-project operational responsibility for maintaining the Sandbox is not adequately described. This is a shortcoming. The quantification of impacts remains unclear. For example, key estimates, such as numbers of patients benefiting, expected gains in diagnostic performance, clinician time savings, and reductions in cognitive burden, are not supported by benchmarks, pilot data, or clear assumptions. For several use cases, including glioblastoma and multi-cancer early detection, the target populations and expected clinical impact are insufficiently quantified. This is a shortcoming. Barriers such as regulatory constraints, interoperability challenges, user adoption, and data-access limitations are generally identified, and mitigation measures are proposed. The governance framework and stakeholder engagement (including patients and regulators) are strengths in this regard. The proposed dissemination, communication, and exploitation measures are generally suitable and well planned. The strategy addresses scientific, clinical, regulatory, industrial, and public audiences through publications, conferences, workshops, digital platforms, and community-building activities. The use of open calls for training and an external user-community programme enhances visibility and adoption potential. The approach aligns well with Horizon Europe’s open- science policy, with explicit commitments to open-access publications and open repositories. The proposal identifies a broad and appropriate set of target groups, including the scientific community, researchers, healthcare institutions, clinicians, SMEs, industry and developers, patient groups, public authorities, European initiatives, policymakers, and the wider public. This broad targeting is appropriate for the project’s scope and intended impact. The management of IP is adequately described. The proposal outlines several potential exploitation routes, including open-source release, licensing options, and service-provision models, and identifies exploitable results with possible patenting potential. Criterion 3 - Quality and efficiency of the implementation Score: 5.00 (Threshold: 4 / 5.00 , Weight: - ) 101290382/BioMedGen-17/02/2026-13:58:49 3/5  Associated with document Ref. Ares(2026)1802897 - 17/02/2026 The following aspects will be taken into account, to the extent that the proposed work corresponds to the description in the work programme: - Quality and effectiveness of the work plan, assessment of risks, and appropriateness of the effort assigned to work packages, and the resources overall. - Capacity and role of each participant, and the extent to which the consortium as a whole brings together the necessary expertise. The work plan is of good quality, coherent, and proportionate to the scale and complexity of the project. It follows a logical structure with a phased approach from development to integration, validation, and open-call expansion, and includes clear deliverables, milestones, and success metrics that enable progress monitoring. The 60-month duration is appropriate for the project’s ambitions. Strengths include robust coordination mechanisms, a balanced consortium with clinical, academic, industry, and NGO representation, and external ethics auditing. However, the Gantt chart is difficult to read, which limits transparency of the timeline. This is a minor shortcoming. The distribution of resources across partners and work packages is appropriate and consistent with the project’s objectives, complexity, and planned activities. Risk assessment is generally comprehensive, addressing data privacy, regulatory compliance, technical integration, and clinical validation. Risk-governance structures, including an independent Ethics and Governance Board, quarterly trust audits, and external validation, are noted strengths. The consortium is well aligned with the project’s objectives and brings together the necessary disciplinary and inter-disciplinary expertise. It includes strong representation from clinical partners with access to patient cohorts and biobanks, academic and computational science groups, AI/ML specialists, biomedical researchers, regulatory and legal experts, ethical and governance specialists, and patient organisations. The inclusion of SMEs and industrial partners strengthens implementation and translational capacity. Overall, the consortium is well suited to the project’s scope. The consortium includes appropriate expertise in open-science practices and in social sciences and humanities, including ethics, societal considerations, legal frameworks, and patient engagement. Gender aspects of R&I are addressed within the consortium. The partners have access to the critical infrastructure needed to carry out the project activities. The consortium demonstrates strong complementarity, combining universities, hospitals, HPC centres, biotech and AI SMEs, research infrastructures, and patient organisations. Roles are generally well defined, and most partners have adequate operational capacity to fulfil their responsibilities. The consortium includes EU industrial developers of GenAI solutions, strengthening the project’s exploitation potential. Scope of the application Status: Yes Comments (in case the proposal is out of scope) Not provided Exceptional funding A third country participant/international organisation not listed in the General Annex to the Main Work Programme may exceptionally receive funding if their participation is essential for carrying out the project (for instance due to outstanding expertise, access to unique know-how, access to research infrastructure, access to particular geographical environments, possibility to involve key partners in emerging markets, access to data, etc.). (For more information, see the HE programme guide ) Please list the concerned applicants and requested grant amount and explain the reasons why. Based on the information provided, the following participants should receive exceptional funding: Not provided Based on the information provided, the following participants should NOT receive exceptional funding: Not provided Use of human embryonic stem cells (hESC) Status: No If YES, please state whether the use of hESC is, or is not, in your opinion, necessary to achieve the scientific objectives of the proposal and the reasons why. Alternatively, please state if it cannot be assessed whether the use of hESC is necessary or not, because of a lack of information. Not provided Use of human embryos Status: No 101290382/BioMedGen-17/02/2026-13:58:49 4/5  Associated with document Ref. Ares(2026)1802897 - 17/02/2026 If YES, please explain how the human embryos will be used in the project. Not provided Activities excluded from funding Status: No If YES, please explain. Not provided Exclusive focus on civil applications Status: Yes If NO, please explain. Not provided Overall comments Not provided 101290382/BioMedGen-17/02/2026-13:58:49 5/5  Digitally sealed by the European Commission Date: 2026.02.17 14:54:53 CET This electronic receipt is a digitally signed version of the document submitted by your organisation. Both the content of the document and a set of metadata have been digitally sealed. This digital signature mechanism, using a public-private key pair mechanism, uniquely binds this eReceipt to the modules of the Funding & Tenders Portal of the European Commission, to the transaction for which it was generated and ensures its full integrity. Therefore a complete digitally signed trail of the transaction is available both for your organisation and for the issuer of the eReceipt. Any attempt to modify the content will lead to a break of the integrity of the electronic signature, which can be verified at any time by clicking on the eReceipt validation symbol. More info about eReceipts can be found in the FAQ page of the Funding & Tenders Portal. (https://ec.europa.eu/info/funding-tenders/opportunities/portal/screen/support/faq) Commission européenne/Europese Commissie, 1049 Bruxelles/Brussel, BELGIQUE/BELGIË - Tel. +32 22991111