New AI platform aims to make biomedical research faster and more precise

Researchers from Aarhus University have helped develop a platform designed to make artificial intelligence more useful in biomedical research.

AI systems have many impressive capabilities, but many researchers have found it difficult to harness their full potential because the systems have lacked access to the specialized databases and tools used in everyday research. 

“For example, our team develops methods to extract molecular information from tissues using microscopy. After we acquire large datasets, we need to understand what our results mean, connect with the literature and propose additional experiments. This is a common and tedious process that applies to most scientific fields”, explains Professor Victor Puelles from the Department of Clinical Medicine at Aarhus University.

“We often spend more time framing the right questions, identifying gaps in the literature, and interpreting results than we do running experiments – and even then, we can’t cover everything that’s out there,” says Malte Kuehl, postdoctoral research fellow in the Laboratory for Complex Tissue Analysis led by Victor Puelles.

“AI can help locate relevant material and integrate findings with the literature, but without access to validated scientific resources and tools, current AI systems can’t deliver reliable results,” Malte Kuehl adds.

Could save countless working hours

To address this, Malte Kuehl and a group of international researchers have developed BioContextAI, a platform to discover plugins that give AI systems direct access to validated scientific databases and bioinformatic software.

Instead of having to “guess” based on its training data, the AI can now retrieve up-to-date information directly from the same databases used by researchers.

Malte Kuehl sees great potential in the approach. For example, when analyzing genetic data and encountering an unfamiliar gene, AI systems can leverage tools from the platform to automatically generate a detailed report.

The system then gathers information about how the gene functions in the body, which diseases it is linked to, whether any drugs target it, and what the latest research says about it.

“Compiling such a report would normally take me hours, but now it can run in the background in just minutes thanks to the tools already available in the platform,” says Malte Kuehl.

An “app store” for research AI

The solution builds on the Model Context Protocol, a new standard that works like an “app store” for AI systems.

“Previous solutions were like the first iPhone – fixed functions but no way to extend them. Now you can install the equivalent of apps into your AI systems, giving them new capabilities as needed. 

BioContextAI’s Registry acts like an app store where researchers can find the tools they need, along with guidance on how to use them,” explains Kuehl.

The project reaches far beyond Aarhus. The team behind the platform includes researchers from Hamburg, Cambridge, Heidelberg, Munich, and several other international institutions.

“Our hope is that BioContextAI will enhance both the quality and scope of biomedical research in the coming years. There are already excellent research tools out there, but they can be hard to use. If we can bridge the gap between practical lab work and computational analysis through AI, we believe the platform has enormous potential,” says Malte Kuehl.

Keeping humans at the center

Despite the ambitious vision, Kuehl emphasizes that AI technologies cannot replace human researchers in biomedical science.

“We’re still a long way from achieving full reliability and efficiency. Research will remain human-centered for the next decade. New experiments still need to be conducted, and hands-on lab work will continue to be an essential part of the process.”

He also stresses the importance of researchers maintaining a critical perspective on AI-generated output.

“Hallucinations can never be completely avoided in generative AI, which is why important decisions must remain in the hands of human experts. All AI systems should be designed to enhance and scale human work – not replace it,” says Malte Kuehl.

A community call

Although the platform already includes a wide range of tools and features, the researchers primarily see it as the beginning of something much bigger.

“We truly believe this initiative serves as call to the scientific community to come together and help shape an emerging ecosystem. BioContextAI provides a unique opportunity to build a resource together that can have immediate benefits in efficiency, scientific quality and overall optimization of resources,” says Victor Puelles. 

“We invite scientists to visit https://biocontext.ai/ to explore available tools and discover how this framework may enhance their research. For us, this represents the beginning of a worldwide effort,” he adds.

Behind the research - more information

Study type: Correspondence / software

Collaborating institutions: Aarhus University, University Hamburg, University Heidelberg, EMBL-EBI, Helmholtz Munich, scverse® community

External funding: Novo Nordisk Foundation, Deutsche Forschungsgemeinschaft (DFG), Aarhus University Foundation, the German Federal Ministry of Education and Research (BMBF), Else Kröner-Fresenius-Stiftung, Open Targets, and Bio x AI Hackathon-2025.

Potential conflicts of interest: Malte Kuehl is employed by and holds ownership in the biotech company KH Biotechnology, which provides consulting services to Lamin Labs - a company that works with data infrastructure for biomedical research.

Link to scientific publication: https://www.nature.com/articles/s41587-025-02900-9

Facts about BioContextAI

• Open-source platform integrating AI with biomedical tools and databases
• Provides access to more than 15 specialized databases
• Developed through an international research collaboration led by Aarhus University researchers
• As an open-source project, all BioContextAI resources are freely available at biocontext.ai

Contact

Malte Kuehl
Aarhus University, Department of Clinical Medicine
Email: malte.kuehl@clin.au.dk

Victor Puelles
Aarhus University, Department of Clinical Medicine
Email: vgpuelles@clin.au.dk