Seven-league boots for precision medicine

Interview

11/28/2024

A conversation with Prof. Andreas Keller from the Helmholtz Institute for Pharmaceutical Research Saarland (HIPS) and Prof. Yang Li from the Centre for Individualised Infection Medicine (CiiM) about artificial intelligence (AI) in individualized infection medicine.

There is a lot of AI in our everyday lives today, for example in search engines, in voice assistants and in various apps on our smartphones. How much AI is there in your research now?

Andreas Keller: (laughs) A lot! AI plays a very central role in our research. Machine learning, which is a sub-area of AI, has been used in infection research for several decades. So the use of AI is not really new here. Neural networks, which in principle function in a similar way to a human brain, are another frequently used form of AI. AI models are valuable tools that enable us to analyze huge amounts of data and extract knowledge and correlations from them. This is also known as data science and is a very important part of our research today. Choosing the right AI model is just as important as the quality of the data.

Yang Li: I completely agree with that! You can't get good analytical results from a poor data basis, even with the help of AI. High-quality, detailed and comprehensive data resources are essential if we are to use AI models efficiently for our research and achieve meaningful results. With modern high-throughput methods, valuable data sets can now be generated faster and more cost-effectively, which we use for our research projects. We also work closely with the Hannover Medical School and therefore have access to large patient cohorts. In my working group, we regularly use AI tools for our data analyses.

How can AI specifically help in individualized infection medicine?

Andreas Keller: What is already being done in some clinics today is to determine the gene sequence of the pathogen in patients with bacterial infections and a severe course of the disease and, with the help of AI tools, derive possible points of attack and receive suggestions for suitable antibiotics. Of course, such rapid diagnostics are not used for every common cold. The process is still too complex and too expensive for that.

Yang Li: But we will certainly continue in this direction in the future. It is conceivable that there will be medical AI assistants in everyday clinical practice that can support doctors on a much broader level in making diagnoses and planning treatment. The emphasis here is on "support", as AI cannot completely replace doctors. One thing is certain: AI will make many things possible in individualized infection medicine in the future that will benefit patients.

Prof Andreas Keller heads the ‘Clinical Bioinformatics’ department at the HIPS in Saarbrücken

Where else could it go in the future? What are you researching?

Andreas Keller: We are interested in the interaction between bacteria and humans. Which bacteria are found in a healthy human microbiome, for example on the skin, in saliva, in the gut? What is the composition of the microbiome when a person suffers from a certain infectious disease? What substances are produced by the bacteria that colonize healthy or sick people? And do these substances possibly have an antibacterial effect?

So you want to find new antibiotic substances!

Andreas Keller: Exactly, we want to use AI to track down new natural substances that could be used as medical active compounds. Against the background of increasing antibiotic resistance, there is an urgent need for research here. But we also want to gain a better understanding of the composition and balance of the human microbiome. Because if we know which bacteria are particularly important candidates for our health, sick patients could perhaps benefit from the targeted administration of customized probiotics, i.e. the "good" bacteria they lack. In future, treatment could thus become more individualized and at the same time gentler. This is because antibiotics, which are always associated with more or less severe side effects, could then be dispensed with in such cases.

Professor Li, what is the focus of your research?

Yang Li: Our focus is particularly on genetics. We want to find out which genetic variants are responsible for the fact that some people have a higher risk of becoming seriously ill with an infectious disease. Or to develop certain symptoms or a particularly high or low immune response. We are also working on models that should make it possible to predict whether a patient will respond to a certain medical treatment or not. Our investigations are based on extensive data sets generated using the latest molecular biology methods, which are summarized under the term "multi-omics". To analyze these multi-omics data sets, we use various selected AI tools depending on the issue at hand, which we adapt and further develop according to the underlying task. Our aim is to identify the key genetic variants that cause patients to be at risk. Based on this knowledge, tailored active substances or therapies can then be developed that can be used as treatment or prevention for risk groups.

Yang Li (left) in conversation with her CiiM colleague Jennifer Debarry

What have you been able to find out in your latest studies?

Yang Li: We were able to identify genetic markers that can be used to predict whether a person will develop a protective immune response after an influenza vaccination or not. By taking these correlations into account, vaccination recommendations could be even more tailored and individualized in the future. The development of alternative vaccines would also be conceivable. The new knowledge that we generate comparatively quickly with AI can advance individualized infection medicine enormously at various levels. Another example is a single-cell multi-omics study with patients suffering from COVID-19. Here, we used multi-omics methods to investigate genetic and epigenetic regulation at the single-cell level. Genetics refers to the instructions encoded in our DNA, while epigenetics involves changes that affect the way these instructions are used. The DNA code itself remains unchanged. By examining the genetic and epigenetic processes at the single cell level, we can see how each individual cell in the body regulates its functions. This gives us a detailed understanding of how our immune system reacts to infections. From the resulting huge data sets, we were then able to use AI to identify molecular markers that could only be found in severely ill patients. We were then able to elucidate the regulatory mechanisms that led to the severe disease. This new knowledge generated with the help of AI now opens up new possibilities and starting points for therapies that could be researched in the coming years.

Professor Keller, you are also looking at the risk of patients developing severe and long-lasting symptoms after an infection.

Andreas Keller: Yes, our aim is to find out more about the so-called post-acute infection syndrome. Many bacterial infectious diseases can be associated with sometimes severe late effects, which often only occur years after infection or after the acute illness. They are then often no longer associated with the previous infection. In cooperation with Alice McHardy's department “Computational Biology for Infection Research” at the Braunschweig Integrated Centre of Systems Biology, we will use AI-supported data analysis to search through existing studies in order to track down the connections between all known bacterial pathogens and documented late effects of infected patients. We want to shed more light on this so that the risk of a post-acute infection syndrome in infectious diseases can be better assessed in future and patients can benefit from better treatment and, in the best case, even prevention. We are focusing our research on the brain, particularly with regard to ageing processes and neurodegenerative diseases such as Alzheimer's and Parkinson's disease. We want to gain a better understanding of the role that infectious diseases can play as triggers or amplifiers. In one of our studies, we were able to show that an infection with SARS-CoV-2 can leave molecular traces in the brain that occur in a similar form in neurodegenerative and other diseases of the central nervous system.

Are there also risks when using AI tools in infection research?

Yang Li: If you simply run data blindly through an AI application and take the results at face value without checking them, that would be negligent and definitely dangerous for later use in infection medicine, yes. It is important to note that the data sets must first be carefully examined, cleaned and prepared for data analysis. Raw data is generally unsuitable here. These preparatory steps mean extra work, but they are absolutely crucial. In addition, the AI tools must be selected according to the research question to be investigated. And last but not least: The analysis results must be carefully checked for plausibility. This means that expertise is required at various levels.

What do you think, is AI the seven-league boot for individualized infection research?

Andreas Keller: Yes, absolutely! Perhaps even the seven^x-league boot! The increase in knowledge that AI has made possible in recent years is truly fantastic. And I estimate that it will continue just as rapidly in the future.

Yang Li: It's a really great time to be doing research into individualized infection medicine. And yes, AI is taking us leagues forward in a very short space of time. The data resources and the AI models available are excellent - and they are getting better and better. So it's going to be incredibly exciting!

Our conversation was also exciting - thank you very much!

Interview: Nicole Silbermann

Prof. Yang Li has headed the department “Computational Biology for Individualised Medicine” of the Centre for Individualised Infection Medicine (CiiM), a joint institution of the HZI and Hannover Medical School, since 2019 and has also been appointed Director of CiiM. Her research focuses on understanding the molecular mechanisms of immunological/infectious diseases through the integration of multi-omics data.

Prof. Andreas Keller has headed the department ”Clinical Bioinformatics” at the Helmholtz Institute for Pharmaceutical Research Saarland (HIPS), a site of the HZI in collaboration with Saarland University, since 2022. Additionally, he has held the professorship for Clinical Bioinformatics at Saarland University since 2013. He researches the interaction between bacteria and humans with the aim of finding new natural substances that can serve as the basis for the development of medical active compounds. Another focus of his work is to better understand the influence of infectious diseases on the brain, ageing processes and neurodegenerative diseases of old age.

This text was published in the HZI magazine InFact in the issue Autumn 2024.