How AI is accelerating the race to understand dementia
"AI will lead to an acceleration in discovery in dementia research that could be transformative for patients.”
Could AI one day help us predict changes in the brain and ultimately prevent dementia? Dr Mathieu Bourdenx, Senior Research Fellow at the UK Dementia Research Institute (UKDRI) at UCL, is using an ‘AI scientist’, which he hopes could transform how we make scientific discoveries.
From automated chatbots on our banking apps to digital assistants like Siri and Alexa in our homes, many of us are embracing AI ‘digital helpers’ or agents to speed up our daily tasks. Beyond the home, AI is also assisting us in the workplace, with almost two-thirds of UK workers reporting using AI for work, such as helping to draft documents or assisting with time management.
But how might AI agents be able to help us in the lab?
Researchers at UCL are exploring this by taking part in a collaboration to test an ‘AI scientist’ with non-profit organisation, FutureHouse.
An ‘AI scientist’ may sound like something straight out of a sci-fi film, however like the AI tools that we’ve all learnt to integrate into our working habits, FutureHouse’s AI platform is designed to enhance rather than replace human scientists.
“An AI scientist is an autonomous AI system capable of independently conducting scientific research; reasoning through problems, formulating hypotheses, planning experiments, and analysing data.” Ludovico Mitchener, Member of Technical Staff at FutureHouse and UCL alumnus, explains: “Rather than replacing human scientists, it performs much of the intellectual labour involved in scientific discovery, allowing researchers to accelerate their own curiosity and explore vastly more questions than previously possible, in tandem with the AI scientist.
“At FutureHouse, we’re building components to tackle each part of the scientific process, from literature reviews and generating hypotheses, to experimental design. These are being developed individually with rigorous benchmarks to assess progress.
“The core of the AI scientist is then an automated loop that executes complex workflows and updates hypotheses and models based on any new data.”
FutureHouse collaborates with leading research groups worldwide, who beta test their tools and provide critical feedback for development.
Dr Mathieu Bourdenx is a member of one of around ten groups world-wide given access to FutureHouse’s latest platform. Dr Bourdenx’s research focuses on understanding the cellular processes that are affected by two fundamental factors: ageing and disease, particularly dementia. “What I am interested in is why, when we age, some brain cells start to go wrong, whilst others are spared or protected from the same damage. If we can understand why brain cells behave differently, this may give us clues in terms of future treatment.”
Dr Bourdenx started collaborating with FutureHouse by testing Finch, an AI data analysis agent. The tool takes in data, primarily in the form of research papers, and allows the researcher to interrogate that data by asking questions, for example “This is a dataset comparing control subjects and Alzheimer’s disease cases. Identify the differentially abundant transcripts/proteins and link the hits to biological functions”. Finch can then provide summaries, including figures, and insights to accelerate understanding.
Talking about the benefit of being able to use AI agents, Dr Bourdenx says: “As any researcher knows, staying on top of the vast and rapidly growing scientific literature is one of our biggest challenges, and Crow, FutureHouse literature-search agent, immediately became an essential part of our daily workflow in the lab. It transformed how efficiently we could identify relevant papers and synthesise existing knowledge.”
Building on this success, FutureHouse recently provided Dr Bourdenx early access to their new data-driven discovery agent to further power their dementia research. Applied to Professor Karen Duff lab’s unpublished datasets (including spatial transcriptomics and single nuclei RNA sequencing generated by Dr Bourdenx and novel proteomics data generated by Dr Martha Foiani), the system identified novel research directions that hadn’t been previously explored. Several findings showed sufficient promise that the group is now conducting laboratory experiments to validate these computational discoveries.
“This rate of progress is quite staggering. Algorithms have progressed significantly, and what I thought would take three years to achieve has only taken six months. With the help of AI, tasks that would take months can be completed in a matter of weeks. Obviously, there is the computational aspect, where AI is able to process vast amounts of data far more efficiently than traditional methods. But what’s really exciting is how AI is now helping with interpretation and hypothesis generation.
“As researchers, we naturally look for what we expect to find based on our training and previous experience. The AI approaches the data more globally,” Dr Bourdenx says.
Traditionally, scientists often follow pathways that make intuitive sense to them based on their past knowledge, training, or simply the tools and expertise available in their particular lab. AI can instantly connect their research to much wider fields and literature than any individual scientist could reasonably master.
“Instead of being limited by the boundaries of our own knowledge or our lab’s specific capabilities, we can pursue the most promising leads wherever they take us,” Dr Bourdenx explains, “This isn’t about replacing scientists, it’s about equipping scientists with more tools to do great work more accurately, and much faster. The AI handles the heavy computational work and highlights patterns we might overlook, but the human scientist remains absolutely essential. We’re the ones who help the AI prioritise which findings are truly significant versus which might be statistical noise. More importantly, we decide what to do next, which experiments to run, which hypotheses to test, and how to translate these computational insights into meaningful biological understanding of aging and disease.”
So where can AI lead us next in research?
“Ultimately, I see AI leading to an acceleration in discovery in dementia research that could be transformative for patients,” predicts Dr Bourdenx, “When you combine human biological intuition with AI’s computational power and remove the bottlenecks that currently slow down research, we’re going to make progress on neurodegenerative diseases much faster than we could before.”