Wednesday, March 15, 2006


Welcome to all Tangled Bank readers! I thought this would be a good opportunity to write some more about our paper (yes, any excuse!) -- in which I explain how our research came about, and drop a lot of names in the process. The point of this exercise is to cut through the neatness (or "fraud", as Medawar famously put it) of the finished scientific paper, and give a more personal account of how we found what we found.

I met Christina Burch, my collaborator on this project, at the 2004 Evolution meetings in Fort Collins, CO. We had a chat over a beer and found ourselves talking about a paper by Siegal & Bergman which showed that selection for any stable gene expression pattern could lead to greater robustness in whatever specific gene expression pattern a genotype happened to evolve. That paper made a big splash in the robustness field by showing that what we might call classical stabilizing selection was not required for the evolution of robustness. It turns out that Christina had thought more deeply about their paper than I had, and had spotted some things that didn't add up. For example, Siegal & Bergman had incorporated sexual reproduction in their model but did not test for its effect. She'd even ran some preliminary simulations and hadn't been able to replicate all of their original results when she removed sex. We agreed that this was something worth pursuing and that we should try and do some work on it together.

Now such resolutions are common at scientific conferences, but they don't always go anywhere. Scientists constantly get excited about new problems, but they aren't always in a position to take on new projects. In this case, things turned out differently for two reasons. First, I was in the process of completing another project that would soon culminate in a satisfying publication (I was actually working on the resubmission during the meeting), so I was looking for something else to do. One direction I was considering (the next logical step, really) was to connect gene networks to our cell lineage models. When I thought more about it, I thought that the Siegal & Bergman model would be a good way to do that. Second, Rolf Lohaus (on the left in the photo) was about to join my lab to do his PhD. He had visited my lab during the Summer of 2003 while working on his Masters project on a gene network model of C. elegans vulval development, and I thought he would be interested in working on it. I was not disappointed -- Rolf started working on the project as soon as he arrived. Soon, two other students joined the project: Suraj Srinivasan in my lab (on the right in the photo), and Kristen Dang in Christina's lab (whom I have yet to meet in person). The work progressed incredibly quickly (the fastest in our experience), a testament to the quality and commitment of the team we assembled: we submitted a manuscript on it in mid-June 2005.

Let's get back to the science. You might be wondering about what happened to Christina's initial questions about Siegal & Bergman's paper. When I reread it more carefully I agreed with her assessment. What did it for me was the result they showed in Figure 2d -- I couldn't believe that it would be true in asexual populations. When we repeated their simulations we found that we were right on that point -- they had actually missed the effect of sex on the evolution of robustness. However, the main finding of their paper (explained above) was still valid in asexual populations (we show that in our paper, although we present our results in a different way).

So why did we decide to look at directional epistasis and the evolution of sex as well? It turns out that both Christina and I immediately saw the interest of those problems. In her case, this will not surprise you. She had long been interested in directional epistasis, as you can judge from these earlier papers from her lab. I suppose doing her PhD in Lin Chao's lab must have played a role in that. In my case it won't be so obvious but it's an interesting example of how formative influences can have far reaching consequences.

Although I had never before worked on the evolution of sex, I've been close to people working on it ever since my PhD days, so I've been quite familiar with the field. I started my PhD in Edinburgh, and the closest lab to ours, both intellectually and socially, was that of the theoretician Nick Barton. The reason why that was significant is that, although I didn't interact much with Nick himself, I did interact with people who were interested in the evolution of sex who passed through his lab, such as Joel Peck and Mike Whitlock. Alex Kondrashov, the father of the mutational deterministic hypothesis, also visited for several weeks while I was there and I heard him speak a few times. When I moved to London with my PhD lab, I continued to hear about the evolution of sex from Daniel Falush, then a PhD student in Andrew Pomiankowski's lab. Throughout my postdoc in Armand Leroi's lab (warning: his lab webpage has been frozen in time for 6 years -- I'm still listed as a postdoc there!), the closest lab to ours (Austin Burt's, again both intellectually and socially), was also actively working on the evolution of sex. For example, Matt Goddard, who went on to do an excellent evolution of sex experiment on yeast, did his PhD in Austin's lab.

Which brings us back to our work. When we started looking at the evolution of robustness in sexual and asexual populations, we quickly decided to check the response in directional epistasis. Our thinking on this was also sharpened by the fine papers on Avida and RNA folding -- in fact, we were practically scooped by this paper on Avida. The rest, as happens so often in science, was luck, but as Pasteur pointed out long ago "dans les champs de l'observation le hasard ne favorise que les esprits préparés" ("chance favors the prepared mind").