The biggest problem in cancer evolution? That mostly people like me are doing it.

Series on Tumor Evolution

Do you know how I ended up working on cancer evolution? I would like to say it was deep thinking, profound insights and scientific vision that led me there, but -alas- the truth is that I was just really bad at proposing interesting projects to one of my first postdocs.

“Evolution? What’s that? Darwin and his finches? Forget it!”

Every time I said ‘How about integrating copy number and gene expression’ or ‘How about reconstructing networks from expression profiles’ he answered politely: “Yes, that sounds very interesting, but I would rather do evolution.”

“Evolution?” I said. “Like Darwin and his finches? Forget it! I don’t think we have that in cancer!” I said. “Be sensible and don’t ruin your career. Do genomics, like I tell you. That’s the future!” I said.

This was in 2009, when I started my group. You can see how little I knew back then. In fact, the only cancer research I knew and did was clustering gene expression and copy number data.

But of course my postdoc was right. He had spotted years before me that cancer evolution was important. And luckily for me, he persisted and today cancer evolution is one of the major themes of my lab.

From genomics to evolution

What finally convinced me to seriously work on cancer evolution were advances in genomic sequencing technologies that connected my data analysis skills with evolutionary questions. I had no eduction in evolutionary theory or any background in anything evolutionary, but analyzing genomic data is something I feel confident about.

And I believe most people working on cancer evolution today are just like me: Genomics people, who somehow ended up tackling evolutionary questions without having been rigorously trained to do so.

This is a problem, because there are many other fields that have something to say about cancer evolution (see the figure below), but either they are silent or they are not well connected to the strong genomics trends. And we genomics people often lack a background in other fields pertinent to cancer evolution (or at least I know I do and suspect most of my colleagues are not doing much better).

bla bla bla
Different fields that all could and should contribute to understanding cancer evolution. Yet, currently most of the action happens exclusively in cancer genomics.

Evolutionary theory

Let’s look at some of these fields, and where else to start but evolutionary theory. A discussion rages at Nature whether evolutionary theory needs a rethink or not, but it is not connected to cancer at all. And other people discuss how to comprehensively link evolutionary and cell-biological thinking — again cancer research doesn’t really seems to be involved.

On the other hand, the people I know in cancer research mostly have no background in evolutionary theory — which doesn’t keep them from reviewing the one or two concepts they have understood.

Population genetics and phylogenetics

Evolutionary theory is the basis of methods in Population Genetics and Phylogenetics. That’s the community of people who publish in MBE and go to SMBE meetings — but as far as I can see there is almost no overlap with the cancer community.

Just look at MBE (Molecular Biology and Evolution), the central journal in this field: Between 1940 and 2013 there have been 58 papers on cancer (or which at least used that word) in 4944 papers listed overall for that period on PubMed. 58 out of 4944 is just above 1%! (I derived these numbers using the code introduced in a previous post.)

Phylogeneticists! Population geneticists! This is to you: Viruses you do, cancer you don’t. Why?

If you are a young researcher in these fields, think about doing work on cancer. Your elders have worked endlessly on fruit flies, Ebola viruses and the like — if you want to step out of their shadows you need to try something new!

And it would certainly be good for cancer research if we had more rigorous population genetic and phylogenetic models.

Mathematical models of evolutionary dynamics

Speaking of rigour, there is a branch of mathematics looking at evolutionary dynamics. Martin Nowak has written a nice book about it.

“Evolutionary concepts such as mutation and selection can be best described when formulated as mathematical equations,” he and Franziska Michor claimed in a review a few years ago.

Maybe. It is not that I disagree. I was brought up as a mathematician, so I know where they are coming from. But as long as these mathematical models are not connected to data they will be just like string theory – fancy but useless.

Our recent review covers both the mathematical models of tumour evolution, like the Moran process, as well as more practical data analysis of genomic data. We struggled how to fit it all together, which showed me first-hand how big the gap here is.

Tumour tissue

Having said this, there is one area where mathematical models seem to get quite some data fodder: the tumour tissue and cellular interactions in the microenvironment. Just look at the very successful collaboration between the Michor and Polyak labs (examples here and here).

But where is the connection to genomics? How do the results on competition of subclonal populations in xenografts help us to understand the data we get from patients, where we have a hard time even estimating the different clones?

And finally: The Clinic

But even if the last point got sorted and the analysis of the tumour tissue was tightly connected to tumour genomics, there is still one major point missing: The clinic.

The main reasons people are investigating tumour evolution (or at least the main reason they say they do) is that they want to help patients. (This is why the arrow between evolution and the clinic is red and faces outwards.)

But except that understanding tumour evolution should somehow be good to understand resistance development which somehow should help us to create better treatment regimes which might help patients –except for this aspiration full of mights and somehows– I don’t see a clear case where research on tumour heterogeneity and cancer evolution has actually helped a patient.

The future of cancer evolution

So where do we need to go from here?

Genomics is currently driving research on cancer evolution but it misses a big chance if it neglects input from all the other fields I have listed who might have something to say on the topic.

And vice versa, other fields should make use of the vast amounts of genomic data that is being collected. You think you know something about evolution and how to model it? Here is your chance to proof it and maybe even help people.

And finally, if we can’t come up with some very robust examples of how research into subclonal populations has benefited patients, the hype will be over much quicker than we all like.


7 thoughts on “The biggest problem in cancer evolution? That mostly people like me are doing it.

  1. yes, we need ‘Evolution’ to learning the dynamics of cancer, and then try to ‘predict’ on the future fate of cancer cell. With the light of Evolution aspect, we can see some thing news at mutli-omics levels. But the traditional population genetics models are two strict, such as coalescent theory(backward inference) and diffusion theory (forward inference). SO, probably we need some new fresh models for complex cancer genomics data.

    Thanks for your posts, I do like them very much, wish to see more thoughtful post from you blog posts and working papers.


  2. Florian–

    Thank you for your very interesting post, especially interesting from the perspective of someone who works on evolution at a cancer research center.

    I do find cancer cell evolution fascinating, and enjoyed learning about your work on MEDICC [1] (we also had a great seminar from Roland [2]) but I do think that there are a few barriers to evolutionary people getting involved.

    First, it’s not yet completely defined what sorts of data should be used. In your work, you decided to use copy number data, which was itself an innovation. One needs such innovation to get good signal, because our old stand-by of point mutations in a few genes aren’t enough to resolve things as far as I can tell. Additionally, I’m not aware of something like TreeBASE for cancer, i.e. a collection of data sets that are suitable for evolutionary analysis.

    Second, it’s clear that there is a lot of work going into defining the actors. I was a little put off looking into the area when there was a recent pile of papers working on various ways to define clones (see recent discussion on phylobabble [3]). I’m not sure if I’m interested in entering such a “hot” field.

    I realize that neither of these should put off enterprising young researchers, but I just wanted to point out that the path isn’t completely obvious. To the extent you can make this more clear, it will become more inviting!

    Thanks again,




    1. Hi Erick,

      thank you very much.

      I didn’t know your Youtube series or Phylobabble (but signed up right now). Which shows again that there is a gap between cancer-evolution and better-understood-evolution.

      And you are completely correct – nobody really knows what a clone is. But why would you want to leave it to evolution-newbies like me to define the terms? Come and help!

      I will continue writing down my ideas here and hope to get some discussion going. ‘What is a clone’ is one of the topics on my todo list.



  3. I believe that there are several good reasons explaining why evolutionary biologists have not been much concerned, with notable exceptions, about cancer. A few of these related just to data might be:

    1) sampling: it is much easier (and I would say funnier) to sample different organisms than human tissues.

    2) little variation: just because of the time scale, the level of variation (=information) observed in cancer is usually orders of magnitude smaller than in organismal (germinal) evolution, making data less amenable to robust statistical inferences.

    3) lack of individuals: evolutionary inference is often made on populations of individuals, or on individuals from different species. Until the appearance of single-cell genomics –which opens a Big Door for evolutionary theory and methods- cancer data was on pooled individuals, which makes, in my opinion, evolutionary inference more complex and less powerful.

    I am sure there are more..but I hope this will change!


You gotta talk to me!

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s