During August and November, statistician Prof. Terry Speed will be touring Australia as the 2014 AMSI-SSAI Lecturer
Can you tell us about a research problem that you’re currently excited about?
Well, I get my excitement from both the science and the statistics that I need to address the science. One of the areas that I’m really interested in these days is called epigenetics. Epi is the Latin prefix – above, on, over.. – epigenetics is really on top of the DNA, giving guidance to an organism to start doing things with the DNA in a cell: reading (transcribing) the DNA, and converting (translating) it to proteins. In some sense, it’s going one step behind what we see. We sometimes see it suggested that DNA is the whole story, but of course there’s a story behind the story, and much of the interest in science is sort of peeling back the layers. So for epigenetics, we typically think in terms of gene regulation – governing what genes go on and what genes go off, and when they go on and when they go off. It’s pretty complicated because we have many important times in the life of an organism and we have many sorts of cells in the body of a complex organism like us. So it’s almost an infinitely complex mosaic to understand. Doing so involves lots of challenging statistical problems and it’s fun biology as well. I probably should say why. Consider the impact of diet. For example, if you feed a worker bee royal jelly you convert it into a queen bee. Diet alone changes the way in which the genes are expressed so that although it was originally sterile, it becomes fertile, it’s able to lay eggs – some pretty dramatic change just from eating some jelly. There are many examples like that.
How are you using statistics to make sense of epigenetic data?
One of the topics that I study is methylation, which is a modification of DNA that plays an important role in expression of genes. It can happen across the entire genome, and from the point of view of an observer you’re interested in finding out exactly where it’s been modified, and how differences of these modifications may play a role. So there’s a general topic of the identification of differentially methylated regions between two sorts of cells. It’s a bit like differentially expressed genes, but as I said, one step behind that. There are lots of challenging issues here. There’s serial correlation in the methylation marks along the genome. That’s quite complex and it mixes in with a lot of genome structure. So, just the simple task of finding which regions are differentially methylated in cell type 1 compared to cell type 2 – there are already a host of interesting statistical challenges in that.
Are there areas outside of medical research where mathematics and statistics are applied that excite you?
Well I get excited about lots of different things at different times, but one of the things that I continue to be interested in is low probabilities. For example, you’ll see low probabilities quoted when people talk about risks. You’ll see low probabilities talked about when people want to say they’re very certain, say of a DNA identification. When you have very little data, as you typically do with these low probabilities, all sorts of interesting and often quite dubious methods got developed to produce answers. So, one interest, not in any sense a major one, is getting to the bottom of what people do and how right or how wrong they are when they are at the limits of the amount of data they have. For example, if you want to say something happens one in a billion times, how do you get data to make that a solid statement?
How important is collaboration in your research?
I think the short answer is ‘very important’, even right back when I was doing my PhD, essentially on my own. Of course because a PhD isn’t supposed to be collaborative, so I used to say that I was a social mathematician, in the sense that I like talking to a lot of people about my work. I’m not that interested in chugging away on my own in total isolation, I like to interact with people, and in a scientific environment it’s a natural thing to do: it’s called collaboration. With mathematicians, sometimes they collaborate in the sense that they both contribute to the same problem, but it’s a lot easier in medical research because you bring different skills to the same issue. In mathematics, perhaps it’s less so – often you have the same sort of skills, and you just carve up the problem in different ways. But anyway, the answer, it’s very important to me because I don’t like working on my own.
Mathematicians and statisticians are often depicted as boring number crunchers. How can we address this image problem?
Well I think the short answer is we just have to get out there. We have to be involved, not just be background people but be involved a little bit more up front. Dive in, don’t be afraid. When you look at physicists or computer scientists, my impression is they tend to be less shrinking violets than we mathematicians and statisticians. But of course we probably think we worry about the details more than big picture people. I’m not speaking against details but you shouldn’t let concern for details inhibit you getting involved in big picture things. Have confidence that we in the statistical and the mathematical world have something to contribute – dive in and then be careful, but don’t let your careful nature inhibit you before you dive in.
As a student, who were your scientific heroes?
I really had one above all others and that was R.A. Fisher, the statistician and geneticist who worked in evolution. I had side heroes (e.g. Norbert Wiener), but Fisher stood out. I did a third year Honours project about his work and I learned a lot about him in my courses. Fisher just stood out. He was known to geneticists as a great geneticist, known to statisticians as a great statistician, known to evolutionary biologists as a great evolutionary biologist. That covered a breadth of my interests as an undergraduate and they’re still very close to my major interests now. There were other people – Darwin, the Huxleys – people who are involved in evolution of course were and are also very important to me.
What attracted you to study mathematics over other sciences?
I think it was basically by elimination. I didn’t find experimental science to my liking. Experiments are great in theory. In practice, I was frustrated – my experiments never quite worked the way they ought to. I think I have an idea of the world – it’s almost a platonic view, that things ought to be perfect. When you do a chemistry or a physics experiment it almost never comes out perfectly. When you do a mathematics experiment, you can get it right, it can come out perfectly. So a desire for perfection, you could say, or if you like a frustration with the imperfections of the real world, led me to this more abstract and more perfect world. Of course I’m also interested in applying the abstract and perfect stuff to the real world, but that is a rather different activity than doing experiments. I think I’m a person who likes science but is a refugee from experimental science.
What are the characteristics you would ascribe to a good teacher?
It probably sounds like a bad thing to say but I think enthusiasm is the number one characteristic. Teachers have to get you interested in what they’re talking about. Of course it’s obviously highly desirable that they’re correct, that they know their stuff, but you could say that for a good student, even a teacher who’s occasionally wrong will get you hooked if they’re interesting. Obviously you’d hope that you’re going to pick up the mistakes if they do have them, but above all, it’s enthusiasm that works for me. Dry, boring teaching and lecturing, no matter how accurate how precise, is not the way to excite and interest young people, at least that’s my view.
You’re a keen advocate for gender equity here at the Walter and Eliza Hall Institute of Medical Research. What do you think would improve female participation in maths-related professions?
Being a male, I tend to focus on the things where the males are possibly at fault or at least things men can do. Clearly there are lots of institutional, structural other barriers to women succeeding in so-called STEM discipline – Science, Technology, Engineering and Mathematics. It’s not just attitudes of men, but men run the show in most of these areas, so it’s a change in the attitudes of men that I think is going to be a major part of changing the situation for women in the STEM disciplines. Being a man and being a supporter of this change, I think I’ve got a role to play. Clearly women have a role to play as well, but it’s not for me to tell them how to live their lives. But if I see men who are being obstructive, who are being conservative or being narrow-minded, who don’t look beyond their immediate male counterparts, I’m very happy to speak up and condemn this sort of activity and trying to move towards a more equitable way of, as it were, involving the entire human race in the activities like mathematics and other disciplines.
You’ve been a researcher for over 40 years, has there been a particular highlight/achievement that you’re most proud of?
I’m never very good with questions like that. Sometimes I look back and think, “Gee, that was kind of nice, I wish I could understand it now” or “How on earth did that idea come?” but usually I’m thinking about the future, not the past. And also of course I am hoping that what I’m going do in the future is better than anything I’ve done in the past. Pointing to something and saying, “Oh yes, that gives me great satisfaction, I’ll rest on my laurels” this is not my style, not something that I want to do. So I’m not going to pick anything in particular. Of course that’s an issue if you’re worried about your, what you might call, your legacy – but for me that’s for other people to decide. I’ll look to the future and hope I can do something in the future that is better than anything I’ve done in the past. That seems to me a reasonable goal.