Biotech company Muna Therapeutics discovers and develops therapies that slow or stop devastating neurodegenerative diseases including Alzheimer’s and Parkinson’s.
Based in Leuven and Copenhagen, it is led by CEO Rita Balice-Gordon who talked to Anne-Marie Demoucelle about the important role that Artificial Intelligence (AI) plays in scientific research, with particular reference to Parkinson’s disease.
Thank you to Dr. Balice-Gordon for her time and insights.
——–
*** How prevalent is AI in today’s research and clinical practice?
AI is everywhere, and we use it in ways we sometimes don’t even realise.
In clinical settings, it is used for example for analyzing blood chemistry, breathing patterns, and EKGs to help predict diagnosis and disease progression. Even common devices like our phones and watches can now generate detailed information about movement, gait, stride length, speed, and balance – all of this data is analysed by AI, often in real-time.
In research — I’ll use the example of Muna Therapeutics because I think we’re quite representative — we’re using AI to analyse human brain tissue in ways that were impossible just a few years ago. We’re examining hundreds of thousands of cells, each expressing tens of thousands of genes differently. It would be like trying to find one molecule of water in a lake – the datasets are simply too vast for any individual to analyse. AI helps us group cells into different types and analyse multiple genetic pathways simultaneously, revealing patterns that human brains can’t see independently.
*** Can you tell us more about Muna’s innovative MiND-MAP platform and how it’s being used?
One of the challenges in drug discovery is translating findings from cells and animals in the lab to efficacy in patients. We’ve developed our MiND-MAP platform to address this challenge. We first used it for Alzheimer’s and recently expanded to Parkinson’s.
We analyse brains from patients who have passed away to understand how cells respond to disease pathology – for example, in Parkinson’s, how cells react to alpha-synuclein and other pathologies. We study both affected cells and those that don’t yet show signs of disease.
What makes our approach unique is that we maintain the geographical information about where cells are located in the brain. Traditional techniques often involve analysing all cells together, which loses the crucial spatial information about different brain regions. Our platform preserves this spatial relationship while analysing gene expression, allowing us to understand what’s happening in specific parts of the brain where neurons are degenerating, such as the substantia nigra, striatum, and cortex. This gives us much richer insights into disease progression and potential therapeutic targets.
*** How do researchers use AI to bridge the gap between laboratory findings and human disease?
We start with human brain analysis to generate initial insights. Then we use AI based tools to analyse large data sets and glean new insights. These are then tested in human cells grown in dishes, where we can manipulate genes to understand their function. Where we can, we then validate findings in animal models. At each step, AI tools help us analyse the results and refine our hypotheses. We’re constantly asking, “Is it the same or different than in humans?” This integrated approach helps increase our chances of discovering effective treatments.
*** Has AI significantly accelerated research timelines?
Absolutely. For example, a spatial gene expression analysis in Parkinson’s that couldn’t even be done 10 years ago, and might have taken two years to complete just five years ago, now takes a fraction of that time. What would have taken armies of people months to analyse can now be done with a few clicks on a computer.
When I was a student, we were sequencing genes using gigantic gels that would take all night to run. Now you can sequence a gene in a machine in a matter of just a few minutes to a half an hour. But when you can gather tons of data, you have to ask yourself: “How am I going to analyse it and what am I going to do with it?” AI and machine learning can help you identify trends and patterns in large data sets that an individual brain can’t grasp by itself. These trends and patterns can be used to generate hypotheses that can then be tested in animal cells, in animal models, in human cells, to accelerate the pace of drug discovery.
*** Beyond research, how else is AI helping in drug development?
AI is revolutionising multiple stages of development. We use AI tools to predict protein structures, identify potential drug targets, and even to draft regulatory documents and clinical trial applications. A regulatory document that might have taken months to draft can now be generated in a couple of weeks using AI tools. At every step, we are using AI and machine learning to help accelerate our work and enhance our chances of success.
*** Are there any limitations to what AI can do in Parkinson’s research?
Yes, there are important limitations. While AI is accelerating many aspects of research, it can’t shorten the actual duration of clinical trials – we still need time to treat patients and assess outcomes. On the other hand, AI tools can be used to predict the patient population likely to benefit, you might end up with a more homogeneous trial population and enhance chances for success. We’re at the cusp of realising AI’s full impact from idea to medicine cabinet, but we’re not quite there yet.
*** What are some common misconceptions about AI in medical research?
One major misconception is that AI is unbiased – in reality, its output depends entirely on the quality of input data. Another is that AI will replace human judgment, but there really is no substitute for human expertise and decision-making. People also sometimes think AI is infallible, but it can and does make mistakes.
*** Could AI reduce the need for human expertise in science and medicine?
I don’t think AI is the biggest risk to the scientific and clinical workforce at all. Actually, it makes doing science, clinical research, and drug discovery much more exciting and compelling because you can obtain new insights faster and those can be developed to have an impact more rapidly. I believe that AI will and already has resulted in better decision making at the research level, at the drug discovery and development level, and clinical treatment level.
The real challenges lie elsewhere – there’s currently less societal appreciation for science and scientists, and the practice of medicine has changed due to infrastructure and economic factors. These issues, rather than AI, are making it harder to attract talented people to the field. We’re seeing concerning trends where admissions and applications are down for medical and scientific careers. This is something we need to address as a society, but it’s not a reflection of AI – it’s more of a post-COVID anti-science rebound that we need to worry about.
*** Looking ahead, what’s the ultimate goal for AI in Parkinson’s research?
Our goal is for these tools to become more powerful and ubiquitous, allowing us to learn more quickly, diagnose patients earlier, and develop medicines faster to help patients at an early stage of disease. If it’s accelerating many aspects of the discovery chain from biology to medicine, it, almost by definition, will result in more shots on goal. More shots on goal mean more goals scored – AI will absolutely have an impact on increasing our chances of success.