We had an interview with Iago Doel-Perez, an enthusiastic researcher with a multidisciplinary background in Data Science, Immunology, and the challenges he overcame while working on his PhD.
Asigmo: Thank you for taking your time to talk to me today. Let’s get right into it with the first question: Could you give me a quick introduction to your professional life so far? What have you been up to and where did you start your education?
Iago: So, I studied pharmacy and biochemistry and also biophysics. I always tried to merge a bit of technology with the application to human health and in the last years, I worked on doing my PhD. There I worked with immunologists, studying the T-cells that nowadays are so famous due to corona. We developed methods to try to talk to them, to try to understand how they react to what they are presented with. For that we used microscopy. I kind of became an expert in image analysis and this is what, more or less, drove me into Data Science to try to extract as much information as possible from the input, from what we get from the images and usually the standard or the very pragmatic problem in biology is that we tend to analyze everything in population-wise, in bulk. Now there is the trend to study everything first on a single cell or molecule level, and then combine that information to do the statistics. And then you get much more information. You get details that you could just miss if you would only look at a bulk.
Asigmo: When did you decide to move in the direction of immunology?
Iago: I felt a bit like a ping pong ball, jumping from biology to biophysics, to the more technical side. As soon as I got in touch with immunology, I got a very good professor that helped me realize how powerful immunology is in many senses. So, I always tried to keep close to it. And finally, I went for a PhD that allowed me to combine all the technology, microscopy and Data Science, etc. Immunology from a strategic point of view is, I think, perfect because it is a system that is not located in one place. Like, you know, you just study the liver, then you know the liver, and that's all you know about the body. The immune system is everywhere. It is very potent, and it is quite powerful. These kinds of therapies, where you can reprogram these T cells and infuse them back to the patient to do a job that they are not naturally programmed for, are now booming. This is for example the topic of CAR-T cells. In therapy, they just managed to cure cancers that were unthinkable to cure. They didn't have any more hope. They were just about to die. They tried this therapy and 80% got cured. And since then there is a boom with like a thousand new clinical essays. I think I just also saw that, the power back then when I decided to join immunology that T-cells are a perfect tool to get into the body and to do these kinds of things.
Asigmo: Am I understanding correctly that you really like both the scientific side and the technical side. How about the human side of helping people?
Iago: Correct. Yes. That's my personal goal, the technology and Data Science and all that are the means, but in the end, I am always going to try to help humans, to improve human health to give opportunities in this field. Of course, by doing research, I am not just trying to help one person individually but trying to break the borders of therapy to reach further away.
Asigmo: Definitely very noble. You've been talking about the technical aspects a bit here and there. When were you first touching points with Data Science during your education?
Iago: So, Data Science per se, I would say very recently, just when I found myself with tons of image data with a bit of knowledge of how to process them and then realizing that I could not extract all the information. There is a lot of it and how do I make sense of it? At the same time, I realized that my problem was not only my problem but a general issue in biology. Now there are these new technologies, for example, next-generation sequencing tools that allow you to sequence the whole genome of a cell or of a pathogen or for example, trying to look at how the cells are responding and which proteins they are now producing. Now we get dumped with data. So realized the bottleneck is now that we have tons of data and don't know what to do with it. I saw also many scientific publications that get the data, they publish something with a lot of figures, but they cannot deliver a message because they don't know how to. There is this gap between possibly the Data Scientists and the biologists. The Data Scientists can provide statistics, but the biologists must understand the meaning and what they can do with it.
Asigmo: A bit into a different direction: I'd like to know if there has been a recent challenge that you've solved in terms of Data Science and image analysis that you're proud of.
Iago: There is going to be a publication and the main topic of my PhD and this was mostly on my own. I have mentioned the bulk analysis before. We have this kind of simple method to know which cells respond. We provide a surface to cells that contain different activating and modulating molecules. We know exactly how many and how they distribute and what they are doing. And then we let cells come in touch with these surfaces. We see if the cells activate or not because we label them so that shows a flare signal or a fluorescent signal when they activate. Until now this technology was used as bulk. One would analyze the whole image with a hundred cells or 1000 cells and get the average. What they did was they individually analyzed all these cells. So instead of getting an average score, I could detect that a population of the cells was actually responding in one way and the other part of the population was responding in a different way. It’s like this thing, they say that a bad statistician says that if your head is in the oven and your feet are in the freezer, you are at the perfect temperature. This is what happened with the cells. Some cells were reacting, and some others didn't. It gives you a totally different message.
I realized how important this data was and then I tried to convey this to my colleagues, my biologist colleagues, and they could not see it. It took quite a while until I could understand why they didn't get it. So, I found a different way to plot it and suddenly for them, it was like a window opened. My project was going in one direction, but because of that, this way of visualizing the data differently and understanding or transmitting the according message, my project shifted into this new direction. Sometimes it's more about just visualizing and seeing it rather than throwing a lot of numbers and tests and hypothesis, null hypothesis and everything else out there. Maybe because I also come from imaging seeing is believing.
Asigmo: Which tools did you mainly use to overcome this problem?
Iago: So, this was before my Data Science course and I used Python tools, NumPy, SciPy, Scikit-image and some tools that were developed by a collaborating lab on single-molecule tracking, single-molecule image analysis. More or less just Python standard tools and Java. Also, I developed a tool in Java, because there is an open-source platform for image analysis, ImageJ, which is written in Java. It offers the ability to add plugins and it is very user-friendly, and it is very well spread. Everybody that knows the word imaging works with it.
Asigmo: Do you think you would approach this problem in a different way now after the Data Science training?
Iago: I think Python and Data Science tools are very fast and they offer very good results. It was in a way a mistake to go into Java as it is not prepared for these kinds of data processing. In the end, you spend a lot of time on things that you can do in Python in one hour. This drives the time spent away from understanding the data, but instead on computing problems. So, I think that's also what I like a lot about Data Science: You can play a lot with the data nowadays on the fly. You can check different hypotheses and it is a bit like a playground.
Asigmo: A very complex playground.
Iago: Yeah. Sometimes yes, but sometimes we tend to over technify it, to make it too complex. And my impression is that we must be a bit like the tale of the naked king. I think current Data Science puts aside all these technicalities in the analysis just for seeing new things and exploring new things.
Find out more about his experience in the Asigmo Data Science Program in Part two of this interview.