AI partner for predicting disease progression and treatment response in brain disorders

About Risorius
 Risorius builds AI systems that take on hard research questions in central nervous system (CNS) disease. A research agent structures the question, reviews the evidence and plans the analysis. Multimodal models trained on patient-level data then return patient stratification and outcome predictions, and the resulting hypotheses go to wet-lab partners for validation. We adapt and fine-tune foundation models to each clinical question and dataset rather than building a new model for every project.

What our models predict

• Disease progression and conversion, for example time to conversion from isolated REM sleep behaviour disorder to Parkinson’s disease or dementia with Lewy bodies

• Treatment response before treatment starts, for example response to antidepressants and to electroconvulsive therapy

• Patient subgroups that share a mechanism, to enrich clinical trials and to re-examine stalled or failed CNS trials

• Biomarker rankings by patient subgroup and disease stage, reported with accuracy, calibration, reproducibility and bias metrics

Data we work with
 Multi-omics (genomics, transcriptomics, proteomics), longitudinal clinical records and real-world data, neuroimaging (MRI, PET), and physiological signals such as EEG, polysomnography and wearable data.

Track record

• IRB-approved studies with 11 hospital partners in Korea, covering metabolic, neurodegenerative, sleep, psychiatric and neurological disorders

• Preliminary results on major depressive disorder and bipolar disorder based on our own clinical data

• Research and proof-of-concept projects with pharmaceutical companies

What we offer a consortium

• The AI and modelling work package, covering multimodal model adaptation, biomarker discovery and ranking, and trustworthiness and bias reporting

• Decision support and visualisation tools for clinicians

• Validation in independent Asian cohorts, adding population diversity to European data (subject to partner hospital approval)

• Federated or in-place analysis, so patient data can stay within each partner’s environment

Partners we are looking for

• Coordinators preparing TOOL-01 proposals in neurodegenerative or other chronic CNS diseases

• Hospitals and cohort studies with longitudinal data on progression or treatment response, especially in Alzheimer’s and Parkinson’s disease

• Pharmaceutical and biotech companies with CNS trial data who want to understand responders and non-responders

Beyond CNS
 We apply the same agent and modelling stack to oncology, where large public multi-omics resources support patient-level prediction studies, and we are developing synthetic patient data and virtual twin approaches. We are open to joining virtual human twin (HORIZON-HLTH-2027-03-TOOL-04) and Cancer Mission consortia.

Topics of interest
 HORIZON-HLTH-2027-02-TOOL-01-two-stage (main), HORIZON-HLTH-2027-03-TOOL-04, HORIZON-HLTH-2027-02-DISEASE-14-two-stage, HORIZON-HLTH-2027-01-CARE-02

Cooperation
 Joint model development, clinical validation, joint publications and Horizon Europe proposals. Data access can follow research agreements, controlled access or federated analysis, in line with GDPR and ethics requirements.