Finnish health-tech SME building machine learning prediction models for dementia risk, multi-morbidity, and complex chronic disease from multi-modal biomedical data.
Combinostics is a Finnish health-technology SME that develops machine learning and data-driven prediction models for complex disease conditions. Their work sits at the intersection of biomedical data science and clinical application: they contribute computational tools and analytical methods to large European research consortia studying conditions like multi-morbidity, dementia, and cardiovascular disease. In practice, this means building predictive models that integrate heterogeneous biomarkers — from epigenetic and microbiome data to lifestyle factors — and turning them into early-warning or risk-stratification tools. Their core value to a consortium is translating messy, multi-modal health data into actionable predictions.
Both EarlyCause and LETHE involve data-driven prediction modelling; LETHE explicitly centres on a personalized prediction model for dementia, while EarlyCause uses causal inference and ML to link early-life stress to multi-morbidity outcomes.
EarlyCause (2020–2024) focuses on causative mechanisms linking early-life stress to depression, cardiovascular disease, and metabolic disorders, requiring integration of epigenetic, microbiome, and clinical data streams.
LETHE (2021–2025), the larger of the two projects at EUR 387,580, builds a personalised prediction and intervention model for early detection and reduction of dementia risk.
EarlyCause lists causal inference as a core keyword alongside animal models and cellular models, suggesting Combinostics contributes statistical and computational causal modelling to multi-source biological datasets.
LETHE introduces lifestyle intervention as a keyword absent from earlier work, indicating a move toward translational outputs where predictions feed into actionable health behaviour recommendations.
In their first H2020 project (EarlyCause, starting 2020), Combinostics was embedded in broad multi-morbidity research — working across depression, cardiovascular disease, metabolic disorders, and using causal inference alongside epigenetics and microbiome data. This reflects a wide-spectrum, mechanistic focus where the goal was understanding causal pathways rather than delivering clinical tools. By 2021, their second project (LETHE) shows a clear narrowing and sharpening: the focus tightens to cognitive decline and dementia specifically, the language shifts from "causal inference" to "data-driven prediction model," and — critically — "lifestyle intervention" appears, signalling a move from research insight toward deployable, patient-facing outputs.
Combinostics is moving from broad multi-disease causal research toward focused, clinically deployable AI tools for dementia detection and prevention — a trajectory that points toward productisation of prediction models rather than basic science.
Combinostics operates exclusively as a participant, never as a project coordinator — a consistent pattern across both projects. Despite the small project count, they have connected with 28 unique partners across 14 countries, which for a 2-project SME indicates they are embedded in large, well-networked RIA consortia rather than small bilateral collaborations. This profile suggests they join as a specialist contributor bringing specific technical capabilities (ML modelling, prediction tools) rather than providing administrative or scientific leadership.
With 28 unique consortium partners across 14 countries from just two projects, Combinostics punches well above its size in network reach — each project it joins appears to be a large multi-partner RIA with pan-European membership. Their Finnish base gives them a Nordic anchor, but their collaborative footprint is fully European.
Combinostics occupies a rare position as a small Finnish tech SME that has been selected as a specialist partner in two substantial European health RIAs — the kind of consortia that typically recruit organisations for specific, hard-to-replace technical contributions. Their combination of machine learning expertise applied specifically to heterogeneous biomedical data (epigenetics, microbiome, clinical records, lifestyle) is narrower and more credible than generic "health AI" claims. For a consortium building a project on brain health, neurodegeneration, or complex chronic disease, they offer a partner who understands both the data science and the biomedical context — without the overhead of a large research institute.
