A Day at the Data Science Spring School
by Mie Tomzak
Last month saw the first joint event between the Danish Diabetes Academy and the Danish Data Science Academy: The Data Science Spring School & Challenge. We talked with the organisers and two participants to hear more about the aim of the event and the experience.
Vejle is a typical Danish town. Not too big, not too small, but quite charming. Combined with wonderfully sunny weather, it makes a great backdrop for gatherings of all sort. From 23rd to 25th of March, a very special gathering took place at the hotel Scandic Vejle. Over 30 early career researchers assembled for three days of learning, task solving, and networking. The Data Science Spring School had commenced, and we were there on the second day to have a talk with the organisers and two of the participants.
The organisers: Adam Hulman & Katharina Herzog
Adam Hulman is a mathematician by training. He has been working with diabetes epidemiology for the last ten years. He works as a senior data scientist at Steno Diabetes Center Aarhus.
Katharina Herzog is lead data scientist at Novo Nordisk. She started as a molecular biologist and then did a PhD in biomedicine, working with patent diagnostics. Her career took her more and more into data, working partly in the field of epidemiology with register-based data, and combining big data, such as metabolomics.
Why make an event such as this?
Adam: There was a gap in the market. The data science and computer science world has its own culture, references, and journals. In the clinical field, it is completely different. Not many clinicians would understand what is going on in computer science, and it’s the same the other way around. We think they could benefit from communicating more, because if you want to solve real-world problems in the health care system, then you should have an interdisciplinary approach.
What are the challenges with trying to bridge data science and clinical research?
Katharina: Communication is definitely a challenge, as well as the lack of knowledge of what the other field does. There is a barrier to break, because one might feel interacting with the other can be scary, so this is a way of trying to bridge the gap and get people connected.
Adam: Yeah, the language is different. And they don’t have machine learning in medical training. They have statistics, and people in clinical research have been using the same methods for decades, but in machine learning something new and exciting comes out every year.
Katharina: And it goes both ways, I think. Take theoretical statisticians for example – some of them never used their work in an applied setting. If you are never exposed to something, it’s very easy to have prejudices. A great example is AI and algorithms, because trust is such a big issue. If you don’t know what an algorithm does, you don’t trust it. If you are never exposed to it, how can you trust a system? So even a little bit of understanding goes a long way.
What are the ideas behind having an event with both boot camp, remote work, and a closing event?
Adam: It’s a three day-course, then remote work in groups for two months, and then there is a closing event, where they present their solutions to each other. One of the main inspirations for the challenge is from data science competitions; here thousands of data scientists compete on online platforms, often using data from the biomedical field. To create more meaningful solutions, both groups have to be represented, both the clinicians and the data scientist. The two months remote project work is for them to get back to their everyday life and their networks, so hopefully they can link to these and use some inputs to learn and think in this new way.
Participant: Sam Lockhart
Sam Lockhart is a trainee clinician at the University of Cambridge. He is a specialist registrar in diabetes and endocrinology, on path to becoming a consultant within these fields. He is currently a PhD student in the O’Rahilly lab in the Wellcome-MRC Institute of Metabolic Science, University of Cambridge. Sam has a long standing interest in diabetes, and is especially motivated by understanding it at a molecular level. His research looks at rare genetic variation, and how it can be used to provide insights into diseases, specifically diabetes and its complications.
Why did you sign up for this event?
Sam: I have seen the DDA host events in Cambridge from time to time. I follow on Twitter and know colleagues who have gone to DDA events and found them very enjoyable. I work in a wet lab environment with experimental biology, mainly, so it’s very hypothesis driven. However, I have been working more and more with groups who use large data sets, and I have been interested in how we can use that in our field. So I saw this event as an opportunity to explore this further. It’s also always nice to step outside your own research system in the UK and visit a place like Denmark, where there is a rich reputation for epidemiology.
What do you think you as a clinician can learn from data science approaches?
Sam: I find that you can achieve real innovation when disciplines meet. As long as we can understand each other and communicate, we can innovate. Other fields have different approaches and methodologies that can offer rich insight, so there is ample opportunity for synergy. The difficulty really is that you need to speak the same language, so even if I leave this course and won’t be able to implement the techniques, I will still have a better understanding of the perspective, the vocabulary, and what may or may not be feasible.
What do you think about this process with a boot camp, remote work, and then a closing event?
Sam: It’s very interesting with this extended format. There are lectures, talks, challenges and group work, and then we have to work at home with our challenges. I think it’s a really great way to do things, because you don’t just come and go, you have to keep working with what you’ve learned, and then you finally see what has been achieved by all the groups at the end of the challenge. This course has a lot of teamwork, and it has been very useful to sit with people with different backgrounds – in terms of everything from nationality to different interests and skillsets.
Participant: Cecilie Cargnelli
Cecilie Cargnelli has just started as a PhD student at the University of Southern Denmark in Odense. With a Master’s in Computational Medicine, Cecilie works in the intersection between biomedicine and data science. During her Ph.D., she will be working on a project on integrative multimodal molecular mapping of the islets of Langerhans at single cell resolution. The goal is to gain better understanding of the pathogenesis of type 2 diabetes, hopefully enabling better treatment in the future.
Why did you sign up for this event?
Cecilie: I have an equal interest in the computer science and biomedical fields, and I find their combination astonishingly interesting and useful. I believe that good communication between these fields is crucial for unlocking the true potential of all the high-throughput data generated nowadays. Right after I had gotten the PhD position, my supervisor sent me the link to this event, and I instantly thought: Of course, this is just my field of interest!
What is your impression of the Data Science Spring School so far?
Cecilie: There has already been so many exciting things, such as the presentations on innovation and on fairness and bias in medicine. You get so many new and interesting perspectives, because we have such a diverse assembly of people here. It’s always great to learn from other people, and I strongly believe that the best results in science are achieved by working together and combining knowledge. This was also a huge motivation for coming here, because it is such a great opportunity to network - especially for an early-career researcher as me, who has only just stepped more fully into this particular area of research on diabetes.
What do you think about the format of the event?
Cecilie: I think it’s a great way of learning, because we also get to work with it from home. In this way, you can attend to your newly acquired network, and you get to maintain the knowledge gained by continuing to work on the group project, and then finish it all with a closing event. There are so many good things to take home from this event.