In this episode, we interview Dr. Michael Love, associate professor in the Department of Biostatistics and Department of Genetics at UNC. Dr. Love earned his doctoral degree in computational biology in 2013 from the Freie Universität and the Max Planck Research School for Molecular Genetics in Berlin.

Listen below!

Highlights

„…the advice I would give to, to myself going back in time is that I think it’s – it’s, you will not appreciate the random interactions you have during your PhD and how those will inform you later on. So, you know go to talks that make no sense to you. Don’t worry about the fact that they make no sense to you. You know, write down the words that people say a bunch and that seemed to be important so you could look them up later.“

„You cannot plan ahead and understand how small conversations […] will be highly relevant for your research program and will be like future directions that you take […] pursue – just like, pursue your curiosity even if you have no training in that area.

from our interview with Dr. Michael Love

Audio

Download or listen to the audio version of the podcast here.

Video

Note: if the player does not load, you can click here to view the video in a new tab/window.

Transcript

Hi.  I’m Michael Love, I’m a associate professor at UNC Chapel Hill in the departments of genetics and biostatistics and I graduated from the IMPRS program in Berlin in computational biology and scientific computing. So I I had three advisors, Martin Vingron and Stefan Haas and Knut Reinert. And then following the PhD I did a post-doc. in Boston with Rafael Irizarry. It was also computational biology bio statistics, and then I transitioned to UNC around six years ago.

Why did you decide to do a doctorate? Take us back to your decision – how did you decide?

So I was I was pursuing a master’s in statistics at Stanford. And there were many different examples of what kind of data could be modeled. And there were many different examples of what kind of data could be modeled. And there were many different examples of what kind of data could be modeled. and I was most interested in the biological datasets so genetic data sets or you know modeling cells and things like this. And I remember I did an internship at UCSF, and there was a postdoc there Owen Solberg And I remember I did an internship at UCSF, and there was a postdoc there Owen Solberg And I remember I did an internship at UCSF, and there was a postdoc there Owen Solberg who talked to me and gave me some career advice that if I wanted to continue this I probably should pursue a doctorate that if I wanted to you know seriously pursue statistical methods for biological data. I should find a a PhD program. 

Why did you choose to do your doctorate at the Freie Universität?

I chose Berlin because in particularly Martin Vingron’s group attracted me because I had seen papers from that group that were right in the area that I was interested in which was developing statistical methods, applying them to new types of data so new data sets that did not have. appropriate – like existing methods that, you know you could just apply. And another thing I noticed was that the methods were being distributed as as software. So you know people could go and use those tools. And apply them to their datasets or, and that those, those tools would undergo some development with you know with other groups. So I thought you know that kind of collaborative process of developing, methods and in close contact with the geneticists and the biologists, and when I arrived in Berlin, I noticed I saw, both with Reiner’s group and Martin Vingron’s group, I saw, you know how close the contact was between the biologists and the computer scientists and the statisticians, that’s what I’d been looking for.

When you started your doctorate, what did you plan to do after graduating? Did you have a clear idea of what you wanted to do?

I had an idea of what I wanted to do but I didn’t know where I would do that. know where I would do that. So you know, and – I had done an internship and during the internship I’d worked closely with people at UCSFs but also there was a team at Genentech. So I had already seen that in pharma there are highly sophisticated computational teams,  statistical teams that were analyzing the exact same kind of data and in many ways had similar questions, and so I’ve thought I recognize, you know I can do this long-term probably in academia or in industry, but I need to get the PhD so I can open the door to either direction. But in terms of what I wanted to do I think I I from the beginning I thought I want to develop these methods and these tools so, I feel like I’m lucky that I so, I feel like I’m lucky that I had that idea and I still get to do that I, I enjoy that I you know I’m still kind of doing a similar thing that I was doing in my PhD which is trying to come up with new methods that people can use to analyze new types of data. And I think the lucky part was that right when I did my PhD there was this explosion of sequencing data sets like sequencing had been  DNA sequencing had been optimized during the human genome project. And the cost had gone down. And then, all of  these different types of data like RNA sequencing or chip sequencing had, you know these datasets were just coming online, so it was a very, it was a great time to, you know, be a computational biologist and trying to like make these methods.

When you think back to your experience with your supervisor or supervisors during your doctorate, what went really well for you?

I had kind of three advisors set up. From the beginning where I was both in the informatics, mathematics and informatics at FU with Knut Reiner’s group and also in the. Max Plank at in Dahlem with Stephen Haas and Martin Vingron and I kind of at the very beginning I alternated between those two campuses and tried to figure out like where, where would I be able to make a contribution? And within the first year I realized that, I was probably going to have a you know more significant contribution in the Vingron department, just not you know I got along fantastically with the Reiner group and got a lot I think I got a lot out of sitting in that group for like the first year or so, but then I realized like, the need for statistical methods and the like – the – in particular these, the smaller groups so group leaders like Ho-Ryun Chung and Sebastian [M..] and, and Peter Arndt, there, there were these group leaders within the department who had very interesting questions and often those had a statistical, thrust to them. And so I was, I, what went really well for me was both interacting with my formal advisors but also being able to talk to the group leaders in the Max Plank and kind of hear from, you know, what are the what are the questions – like – Sebastian, what are the questions that he has about transcription factor binding sites or from Ho-Ryun, like, what are the questions with epigenetic regulation that they were pursuing in their group. And how can statistics be used to help answer those questions. So. The fact that it was a, it was a big department, it was very diverse in terms of the, the fields. And there was a lot of kind of, cross-pollination.  So being able to both, you know hear what’s exciting and then So being able to both, you know hear what’s exciting and then hear what’s challenging and what what’s needed to, you know, what, what would be useful to have a tool to do XYZ? Or also like, oftentimes the first like really interesting new experiments producing new types of data, it takes a lot of Iteration between the computational side like it takes a lot of Iteration between the computational side like pulling up the data starting to visualize it and the experimental side saying like, well oh by the way there are these, you know, technical biases we need to worry about, or, what we’re really looking for is this signal here at this you know place in the genome. So having those teams. You know physically close to each other and so you can interact with each other often is really helpful. It speeds up the development. Whereas if you’re on separate camp- campuses you can still, you know have that conversation but it takes a lot longer to iterate. And so, that was a very productive environment for me. You know, Martin, as the head of the department cultivated that, I think he, you know, he brought these people together. He encouraged them to interact with each other. There were rotating presentations in the weekly meetings. And so, you know you were able to, very quickly like hear about a new type of data offer to look at it and interact with a PhD student from the experimental side. And iterate and then see what we could figure out.  

Is interdisciplinary communication common in your field? Was this something unique about your program?

Yeah I think that so my from my experience the, the Berlin program. managed to do this well. It’s difficult to foster interdisciplinary work because you’re often on the edges of institutions. So institutions, you have departments, and within a department there are recognized types of work that you could do and types of scholarship – and interdisciplinary work, you’re by definition you’re on the edge. And so of a lot of universities tried to, you know create interdisciplinary, try to reward interdisciplinary teams or like they, you know, that’s a buzzword that that universities and research institutes want to foster but it’s another thing actually accomplishing it. And so yeah, that was really really helpful for me really impactful for me. And I’ve tried to seek that out everywhere I’ve been since. I’ve tried to like, I want to find a place where there’s a lot of crosstalk between, you know, doctors geneticists biologists and the computational side.  That was part of the, IMPRS. So the international Max Plank research school. that is, I think it’s, it’s like it’s housed between the Max Planck Institute for Molecular Genetics and Freie. And also has lots of other participating institutions and PI and investigators so that was so that was the degree program that I received my degree from. And it’s kind of, it’s tied to, it’s part of the Bereich mathematics, informatics at FU. So like that was where I had to go to get my diploma, but the program was this IMPRS program.

You’ve been in the US, where programs might be more structured, and in Germany, where it is less common to do coursework – can you tell us a little bit about your experience in both systems?

So I had you know I I think I had a strange experience because I, in my master’s in statistics I You know, I got all this coursework and but then no you know I did I wrote I did a research project, but I didn’t you know have a dissertation out of the masters. And then I transitioned to Berlin where I took some classes but you know predominantly just research just working on you know, research that will lead to first author publications is kind of the main goal. And so there was this hard split, like ocean, in between my coursework and then my research in the PhD. And so you know that, that was I think I, I benefited from that because I had these tools, these, these statistical and mathematical tools for thinking about data. And then it was just like, hit the ground running and start to apply them. There was, a rich, a culture of like keeping up with the, statistical literature or like mathematical literature. So oftentimes in group meetings we would be talking about, you know, papers that had just come out, but the focus was on new methods and kind of assuming that you had a baseline of, you know, statistical methods if you were going to be developing new ones. So we could read this new paper and you knew what they were doing, like what what the stochastic model is underlying that. So I thought you know it worked out well for me.  One thing that I, I didn’t experience this in my labs but like something I’ve seen you know, in my travels around the world is that you can end up – in particular people who are focusing on computational work you can end up in labs where you don’t have the theoretical background or mentorship that you need to succeed. Where, because it’s so desirable to have someone who’s capable at, at, data science or writing algorithms you can end up, like getting accepted to a position where you will not get the mentorship you need. And so that’s something to be concerned with. I didn’t, I didn’t experience that, I I felt like I always was able to receive whatever mentorship I needed by seeking it out from, you know, various experts around me, but like that’s a, that’s a, I think a danger of not having like, the ground, you know, not having like everyone should take these courses in order before they begin dissertation work. Not really for me, it was a really good setup I received. I can still remember these, like, soft skills courses from IMPRS and I kind of, those kind of filled in all the, all the gaps that I can think of. Like, I still remember there was a writing soft skills course that I think really influenced the way that I still write manuscripts and the way that I teach students how to write, like write for an international audience and you know, lot lots of like great practical advice like that. About scientific writing. And we were also encouraged to, you know a lot of the students in the lab would, as part of their PhD also like have a short visit at another lab where you could see you know that – that’s not that common. So I was very grateful that I was able to do that. So that was you know, part of my PhD was I visited EMBL in Heidelberg, which was a really productive time in my PhD to kind of see what they were doing. In Wolfgang Huber’s group there.

When you look back on your experience – with your supervisor, with your dissertation – is there anything that you wish had gone differently?

 I could say one thing there which is kind of a universal advice, is that there’s this – there’s this blog post from Floren Markowitz. who was a computational biologist in the UK and he writes about, really you should conceive of your relationship with your advisor as you’re not being managed but you’re kind of also managing them. It’s – it’s absolutely a two way relationship and you cannot, you should not just expect that, you know, they will, that that you’re, it’s not that you’re doing work for them but really they’re, should see them as a resource for you. And it’s a finite resource. Both in time, like you will only have access to this person for the amount of time that you’re in their lab, and also like they’re split among other students and post-docs and so you want to make the most of of that finite resource and you know so yeah, thinking critically about how about the engagement that you do have with your advisor as a point of like things you’re managing, you’re managing up or what – you know, you don’t want to make it up or down, you’re managing them as well.  It’s really good, and I send it now to everybody. Okay so one thing that I’ve tried to reproduce is, is pairing up computational or statistical students with the, you know, people that are doing the experiments as much as possible. And like, have it, like also taking the advisors out of the loop. You know if there’s two advisers, say, so that they can, interact and ask questions directly. Because that’s, that’s how that’s how you really figure out what’s going on and all the details. Because a lot of times, you know the advisors might not know all the you know, like day-to-day details about you know datasets and oh you know, this new experiments over here in this so trying to pair up computational with experimental teams. And then, you know just having that interface. Rather than, having the experimental data just be like, especially, there’s a lot of publicly available data, which you can just download but then you really don’t get the important information about like why was this generated? What was the questions being asked? You know what is the positive control here, what’s the, what do we expect to be positive what do we expect to be negative, all that really important metadata about the experiment is not there, but you get that If you have a on-site collaborator.

As a professor, you’re currently supervising PhD students and you’ve supervised students in the past. What’s something that you took with you from your own experience for your students?

So trying to provide a setting where they will, where students will have interactions like across, cross field. And so they can learn to speak a different language. Like learn to understand the language of a biologist or geneticist. And, you know there’s a lot there’s an abundance of publicly available data and it’s very you know of course useful and you can write lots of papers with publicly available data, but like, I think you miss out if you just work exclusively with the public data because you never get to hear those questions that motivated the generation of the data sets. So that’s something I’ve tried to recreate in my lab. So I did not, from my PhD I was I was not certain, graduating from the PhD program that I would, you know, definitely want to stay on in academia. And I remember. like in my post-doc, talking often with leaders of computational teams in pharma. because throughout, from my master’s to my PhD till now I still, I interact a lot with very sophisticated teams that are doing fun innovative research, developing tools and methods in industry and so I never thought that you know this would only be a this kind of like interdisciplinary teams and making new methods and writing papers, going to conferences – I never conceived of that as something that would be only available in academia. I’m, I’m happy in my current position, I, I really like working with students and post-docs and I, when I – when students asked me like, you know what’s the difference I think a major difference is whether you enjoy training. Whether you enjoy you know, taking somebody who’s never seen genomic dataset before and training them to when they, you know, when they defend their dissertation versus you can do similar work in industry with people that are trained from day one. So like that’s kind of for me the major difference. I think. I recognized during my postdoc that it would be possible for me to seek a investigator position, like a professor position. In departments in the US, that, you know, there’s, there are cycles to the academic job market and I was lucky to come in at the right time, if you look at you know, between recessions right? Like there’s a lot of luck involved and so it was a good time. There was a lot of interest in people who had data science skills or, people who were trained in computational biology and genomics. It was just a lot of interest in, in the departments kind of building that out. And so I could tell that it would be an option for me to go on the academic job market, it’s, you know, but it does change year to year. And another thing that kind of, another thing that, that I used as part of my job search was that throughout from PhD to postdoc to to my position now I used a network, which is this open source software project called Bioconductor. So, I started my PhD So, I started my PhD working for the, working as part of this big collaborative project. I used that to meet people. That’s how I met my postdoc advisor. And then that was also that, that project helped me helped like elevate my work to a level that when I was on the job market people were familiar with the papers that I’d written. I think primarily because I’ve been putting out the software on Bioconductor, which is like a platform for showing off your work. And that’s a specific platform for a specific type of, you know, set of methods. But if you could find that. Like in your field if you can find some international you know, collaborative way to show off your work. Then that’s, you know, that – that can help you like bridge these gaps because then people will already know what you’ve been working on at each step in the process.

How did you know that you wanted to stay in academia? Did you consider a different track, or did you always know that you wanted to stay?

That had been built for me. Right, so like that existed, that network existed. It was started in 2004 or something. I graduated – I was in my PhD 2010 to 2013. So that existed, I could just jump onto that, other people like they might have to create their own network and that’s a lot more work. And like, you know that, it just was, it happened to be the right network for me. It was built top of this language that I had specialized in my master’s program I you know, I was familiar with that programming language. And it was about the data that I was interested in analyzing. So it was just like perfectly built for me to you know, put my work on there and then have it be seen by others, see. You may have to, for the first one, you may have to build the network yourself from scratch and that’s just a ton of work.

From where you are today, what would you tell someone who wants to go into academia? Is there any advice that you would give?

So I think, there you. for your, for your own, like emotional wellbeing it’s really important to recognize that there’s a lot of luck and cyclical nature to academic jobs. And, so departments go through phases. They have capacity to take on new new positions and they, at the time will, they’ll conceive of like this is what we need. We need someone who can do this. We need somebody who can do machine learning and look at image data, or we need somebody who can do statistical genetics and look at rare variants you know so it’s, I It’s not always so specific but, but It helps to have those, it helps to have already, done work in those areas and that, and there’s a, there’s an aspect of like luck. If you happen to be, you know, have that in your CV. So one thing that you can do to, you know, to – to plan for that is to try to have some diversity in the papers that go into your dissertation or the chapters of your dissertation. Like the broader you spread yourself the more chance you have that something on there will appeal later on to, to a committee, a search committee. So you know, something here in, in cancer biology and something here in, in gene regulation yeah so that’s kind of specific to my field but, kind of, you know, having a little bit of spread in the chapters of your dissertation will help you like attract a postdoc advisor and then, and then appeal to a search committee later on. But it’s not like, you know there’s a – if you ask people who have positions, like, what did you do? There’s a lot of, ascertainment bias in that. Like, you know they’ll tell you everything that they did, but you’re not hearing from, you know a representative sample – as a statistician I have to bring up the the bias problem yeah and then I, you know, I don’t know much about, I can’t I can’t comment much because I didn’t – I never sought out a position in Germany. as a PI. So, yeah, I, yeah I don’t have I don’t I wouldn’t be able to compare very well.

Our last question – if you could time travel back to the beginning of your doctorate to give yourself advice, what would you say to your past self?

Sure. so the advice I would give to, to myself going back in time is that I think it’s – it’s, you will not appreciate the random interactions you have during your PhD and how those will inform you later on. So, you know go to talks that make no sense to you. Don’t worry about the fact that they make no sense to you. You know, write down the words that people say a bunch and that seemed to be important so you could look them up later. And, you cannot – you can’t plan ahead and understand how small conversations you have at a conference or talks you go to that are totally weird will later on, you know, be highly relevant for your research program and will be like future directions that you take. So there’s a, there’s a there’s an element of, of like spontaneity that you should pursue – just like, pursue your curiosity even if you have no training in that area. Just you know go to those weird talks and sessions.