What We Do
AIRIS is developing an AI collaborator for biomedical research – an AI-assisted environment designed to support researchers across the entire scientific discovery process. By supporting automatic data integration and processing, mechanistic reasoning, hypothesis generation and study design, AIRIS can help researchers move more efficiently from fragmented data to testable scientific ideas. This streamlines the research process and allows greater focus on scientific interpretation and discovery.
AIRIS in Action
What could working with AIRIS look like in practice? Imagine Elena, a researcher investigating pulmonary fibrosis, who wants to understand why the disease progresses rapidly in some patients but more slowly in others. Click through the process.
- 01
Harmonising the Research Data
Elena starts with data from different sources, including medical imaging, lung function measurements, omics data and clinical records. Before she can investigate her research question, these datasets need to be brought together and prepared for analysis.
AIRIS supports her by harmonising the different data sources, aligning information and identifying missing data. Where appropriate, synthetic data can help address gaps. This reduces time spent on data preparation and gives Elena a more consistent basis for further analysis.
- 02
Exploring What Drives Disease
With the data connected, Elena can investigate why disease progression differs between patients.
AIRIS combines the available data with causal and mechanistic models to explore relationships across biological scales, from molecular and cellular processes to patient-level outcomes. Elena can also investigate “what-if” questions, such as how a patient's disease trajectory might have changed under a different treatment or intervention.
- 03
Turning Insights Into Testable Hypotheses
Based on these analyses, Elena asks AIRIS to identify promising explanations that could be investigated further.
AIRIS combines the data with scientific literature and biological knowledge to propose testable hypotheses, for example, potential biomarkers or therapeutic targets. Specialised AI agents examine their supporting evidence, biological plausibility and novelty before Elena reviews the suggestions herself, refining promising ideas and rejecting those that are less convincing.
- 04
Taking the Next Step Towards Validation
Once Elena identifies a promising hypothesis, AIRIS helps her plan how it could be tested.
The ΑΙ collaborator can support decisions around study design, participant criteria and statistical analysis and help turn the research question into a draft study protocol. Elena can then review and refine the proposed study before moving towards experimental or clinical validation.
Behind the AI Collaborator
Behind the AIRIS AI collaborator is a family of specialised AI agents that support different research tasks. Working together, they can analyse information, generate ideas and critically review outputs, helping connect different stages of the research process while researchers remain actively involved.
Explore the agents behind AIRIS
Hover over or select an agent to discover how it supports the AIRIS AI Collaborator throughout the research process.
Data Integration
Schema linking and homogenisation agents help connect heterogeneous biomedical datasets and bring them into consistent representations.
Data Analysis
A data analysis agent supports exploratory analyses and reporting, while allowing researchers to define analyses for their own research questions.
Synthetic Data Generation
A specialised synthetic data generation agent can help address data gaps, support cross-modal imputation and create data for simulations and further analyses.
Hypothesis Generation & Review
Specialised agents can propose ideas, review literature, validate hypotheses against data, and assess their novelty, plausibility and supporting evidence.
Mechanism-Informed Review
Reviewing agents can compare outputs with underlying biological and mechanistic models, helping identify inconsistencies and strengthen the plausibility of results.
AI Supports. Researchers Decide.
AIRIS is designed as a research support tool, not a replacement for scientific judgement. Researchers remain responsible for reviewing results, refining analyses and deciding which hypotheses to pursue. To support this critical assessment, AIRIS will provide supporting evidence, mechanistic explanations and information on uncertainty to help them critically assess its outputs.
More information on explainability, fairness, robustness and human oversight can be found under Trustworthy AI.