Nia Cason: Hi Félix, could you introduce yourself in a few words?
Félix Balazard: Hi, I’m Félix Balazard, and I’m a PhD student at INSERM and Sorbonne Université (Paris). I started my PhD in 2015. I’m at two different sites, which means I have two different PhD supervisors; Gérard Biau, who’s a statistics professor at Sorbonne Université, and Pierre Bougnères, who’s a medical doctor at INSERM and works on type 1 diabetes. My PhD is about the causes of type 1 diabetes and personalised medicine, from the perspective of applied mathematics.
NC: What inspired you to choose this research topic?
FB: My background is in mathematics, so I studied that at the École Normale Supérieure, and then later I became interested in biology. I didn’t want to stay in purely theoretical mathematics – I wanted to see the impact of my work in a more direct way, so I decided to go into a more applied field. So in my case, the application is medicine; I’m trying to apply statistics to medicine to try and figure out what causes disease
NC: Could you describe your current research?
FB: Right now, I’m writing my PhD thesis, but I’ve written maybe four articles and one has been published so far. It’s a paper on type 1 diabetes that’s been published in BMC Public Health and is about the analysis of a large environmental questionnaire on type 1 diabetes: “Association of environmental markers with childhood type 1 diabetes mellitus revealed by a long questionnaire on early life exposures and lifestyle in a case–control study”.
The last paper I worked on is on personalized medicine. The way medicine usually works is to take data from a randomised clinical trial, look at the average treatment effect, and use that to identify the ‘best’ treatment. We developed a methodology to see if you can personalise treatment using the same set of clinical data. So maybe one treatment is indeed better on average, but maybe, for some people, another treatment is optimal. So, using the individual characteristics of patients, you can predict which treatment would be best for them. The bigger picture of this work is personalising treatment according to a patient’s individual characteristics, so it’s pretty exciting.
NC: Does applied mathematics + biology have anything to do with… Westworld?
FB: I think there’s quite a gap between the public perception of machine learning and what we’re actually able to do. There are very impressive things, such as image recognition, but it’s very far from being intelligent in any way. Of course, there’s lots of philosophical questions that people are most interested in, more so than what machine learning is actually about, and that’s normal, but the reality is that we’re very far away from those questions.
There are other more immediate questions, such as fairness – such as the fact that discrimination can be reproduced by algorithms. Take predictive policing, for example, where you give an algorithm the places where arrests have been made, and it predicts where a crime is likely to be committed – so, the police arrive to that place, maybe they find something maybe not, but it’s going to reinforce what’s already in the data. This is relevant to the over-policing of black neighbourhoods in the US, for example – the notion that an algorithm is ‘fair’ is false, because the initial input data were prejudiced.
NC: What problems have you encountered during your thesis?
FB: I’d say there are two main things. The first is that you can have an idea and you think it’s great, but going from this initial idea to its actual implementation and then writing a paper can take months, sometimes years, so, in reality, the end result can be quite far away. The second thing is that I work alone a lot. This means you really have to motivate yourself. That’s why I joined Doc’Up, the PhD Candidate Association at Sorbonne University, which has things like scientific outreach activities, social activities, and post-PhD career advice.
Also, a PhD is kind of overwhelming – even though you can sleep in late and go into the lab in the afternoon if you like, you’re always thinking about your work – it’s overwhelming in the sense that it becomes very central in your life.
NC: What do you plan to do after your PhD?
FB: I’m going to continue research in the private sector. I’ve already found a job at Owkin, which is a start-up company that works on machine learning applied to health, so this is very closely related to my PhD. I’ll start in September.
I was lucky to find this position – I had a friend of a friend who worked there, so got an interview through these contacts. Also, I’m in a very popular field right now, artificial intelligence and health, so the job market for me and others in my field is very good – but you do need to start thinking about what you’re going to do during your PhD rather than after it.
NC: What scientific events do you recommend?
FB: My subject relates to many different areas, including genetics, medicine, and statistics, so I’ve had the chance to attend some very different and enriching conferences. For example, The Gordon Research Conference on quantitative genomics, that was really great, with the top people in the field. And a great location, too – a luxury hotel in Tuscany!
I also went to the Journées de Statistique during my PhD, which is a French statisticians event with a real sense of community. There’s also some shorter conferences, such as the Statistical Model for Post Genomic Data, which is an annual two-day conference. Also, in Paris there’s lots of seminars. I actually stopped going to many because otherwise you don’t do anything else with your week!
NC: What scientific personalities do you follow online?
FB: Actually, there’s some very interesting science being exposed on twitter; for example, there’s Sek Kathiresan who’s a cardiologist involved in some very interesting genetics research in the States. There’s also quite a big twitter community in the field of machine learning and artificial intelligence, so I find Francois Chollet a great read.
There’s two blogs in particular that I like – there’s the blog of Andrew Gelman, who’s very good on the replication crisis – he goes after ‘bad’ scientists and bad use of statistics. The second is Lior Pachter’s blog. He has a maths background and is working in biology. He’s also known for his controversial style, attacking colleagues and using all kinds of … adjectives!
Researchers or not, is science and/or tech your job or your passion? Share it with the community!
To do so, contact us directly: [email protected]