Interview | Prof. Dr. Peter Nicholas Robinson "Humboldt Professor for Artificial Intelligence 2024"

Using data for precision medicine

Bioinformatician Prof. Dr. Peter N. Robinson is a pioneer in the computer-assisted genome and phenotype analysis of genetic diseases. The Human Phenotype Ontology he developed is now a standard tool used internationally to diagnose gene-related diseases. He has spent a large part of his research and teaching career in Berlin. After several years in the US, he returned to the German capital at the beginning of the year as Alexander von Humboldt Professor of Artificial Intelligence at the Berlin Institute of Health at Charité. We spoke to him about his view of Berlin since his return and about his research. 

1. What motivated you to return to Berlin? 

The Berlin Institute of Health (BIH) offers the opportunity to conduct translational research in Germany at the highest international level. This is because the BIH has succeeded in building an effective bridge between research and day-to-day practice in clinics. Our aim is to test and advance artificial intelligence and algorithms in close proximity to clinical practice, which is comparatively easy to do here at BIH as part of the Charité. I am already liaising with clinical groups at the Charité to collaboratively test the application of AI in medical research. This proximity to clinical practice, the associated opportunities and the excellence of the research were the main reasons for my return. Ultimately, Berlin is also an ideal location to conduct research and it's also a good place to live. 


2. Can you briefly explain the Human Phenotype Ontology (HPO) that you have developed? What questions can it answer in medicine? 

The HPO is an ontology of genetic-associated diseases. In other words, it is a database that assigns clinical manifestations of diseases to the corresponding gene mutations and syndromes. The challenge we are trying to master with the ontology is to enable computers to analyse human knowledge. To this end, an ontology structures knowledge as hierarchical data structures. With the HPO, we have specified around 17,000 symptoms as terms - so-called "ontology terms" - for the computer in such a way that they can be used for artificial intelligence (AI). A simple example of such a specification is the linking of a disease such as cataracts in the HPO with peripheral cataracts. In this way, the computer knows that these diseases belong together. In parallel, there are disease models that are made up of symptoms of diseases and other information such as genomes. In this way, a semantic network of terms has been created that can improve diagnostics with the help of AI. 


3.Which future collaborations, projects and initiatives in the region and in Germany are you looking forward to most? 

We have already decided on a specific collaboration with clinical groups at Charité. As this will start soon, it’s what I’m currently looking forward to most. In addition, there will be future collaborations with many groups in Germany, Europe and worldwide to further expand the HPO. This will also always involve fulfilling the requirements within the respective field, which in itself is an exciting task. In this context, I will also expand cooperation with the European reference networks for rare diseases, which I hope will lead to great progress. Basically, we are lucky to have any number of promising and exciting collaborations.  


4. What do you think the medicine of the future will look like? And what role will AI and big data play on the path to personalised precision medicine? 

Put simply, a doctor today has various options when treating a disease. The diagnosis identifies the disease and the treating physician chooses the therapy that works best for everyone on average. However, it is safe to assume that there are subtypes of most of the currently known diseases. The approach of precision medicine is to precisely identify these subtypes, to adapt the treatment to them and thus to tailor it to the respective patient, because a specific therapy is usually more successful. This is now a common approach in the treatment of breast cancer in women, for example, and has already saved many lives. However, with a few exceptions, this is not yet the case for other diseases. I believe that in the coming decades, AI will make it possible to better understand the subgroups of diseases and thus apply precision medicine to most diseases. Both AI and big data will be used extensively to make this process possible. 


5. How important are the BIH and your research focus for the Berlin AI ecosystem and the HealthCapital healthcare industries cluster? 

The colleagues at the BIH are performing research at the highest level, and this also applies to my research area. I think that the BIH, with its research in the Berlin AI ecosystem, can be one of the drivers for the broader application of AI in medicine. I think we will also be able to make a contribution beyond that. 


Brief bio: 

Peter N. Robinson began his academic career studying mathematics and computer science at Columbia University in New York. He then studied medicine at the University of Pennsylvania in Philadelphia. His path then led him from the US to Germany. In 2000, he completed his specialist training in paediatrics at Charité – Universitätsmedizin Berlin. This was followed by a switch to human genetics. During this time, he completed a master's degree in computer science at Columbia University via distance learning, which enabled him to go into bioinformatics. At Charité, he then held a professorship at the Institute of Medical Genetics and Human Genetics, while also serving as a fellow at the Berlin-based Max Planck Institute for Molecular Genetics and a co-opted professor at the Institute for Bioinformatics at Freie Universität Berlin. In 2016, he moved to the Jackson Laboratory for Genomic Medicine in Connecticut as Professor of Computational Biology. After seven years, he returned to Berlin in early 2024 with an Alexander von Humboldt Professorship for Artificial Intelligence at the BIH. 


Further links: