Picture: Alois Endl
This is the first one from a planned series of occasional interviews of The Reference Frame with some of the most distinguished very young scientists (below 30 or so). It is likely that many of the interviewed people are going to be Harvard Junior Fellows; however, you may suggest your own candidates.
Update: See also Esquire about Franziska MichorQ: We will start with Franziska Michor, who is a Harvard Junior Fellow working on mathematical models of cancer. Franziska, can you quickly describe how you got here?
A: I did my undergraduate in molecular biology and mathematics at the University of Vienna, during which I spent half a year in Italy studying medical biotechnology. After that I started my PhD at Harvard University with a sabbatical at the IAS in Princeton. I got my PhD from Harvard in 2005, and now I'm a Junior Fellow.
Q: When you study cancer, do your results apply to all types of cancer or is there a difference between colon cancer, lung cancer, or leukemia, among other examples?
A: Some of my work is concerned with genetic alterations that occur in all kinds of cancers. For example, tumor suppressor genes are genes that contribute to cancer progression if both copies are inactivated or lost from the cell. There are many different tumor suppressors that are involved in different types of cancer, but they all function similarly inasmuch as they need to be inactivated in both copies for tumorigenesis. Therefore, an interesting question for most types of cancers is how long it takes to inactivate a tumor suppressor gene in a population of N cells. The answer represents an important step towards a quantitative understanding of cancer genetics. Another such example is the role of genetic instability in tumorigenesis. Other projects, however, are concerned with mutations or treatment strategies that are specific to one kind of cancer.
Q: What are genetic instabilities, and how did you use mathematics to investigate their role in cancers?
A: Genetic instabilities come in two different forms: microsatellite instability is an increased point mutation rate, i.e. an increased probability of changing single base pairs in the genome; chromosomal instability is an increased rate of losing or gaining whole chromosomes during cell division. Genetic instabilities can be detected in 100% of late-stage cancers. One of the most hotly debated questions in cancer research is whether genetic instabilities arise early in tumorigenesis and thus drive most of the mutational accumulation in a cancer cell. This question is tricky to answer experimentally, but a mathematical model could establish that genetic instabilities are very likely to initiate tumorigenesis. This result has consequences for the basic understanding of cancer progression as well as for treatment strategies.
Q: You mentioned your work on specific types of cancer. Can you tell us more about that?
A: I've done some analysis on chronic myeloid leukemia (CML). CML arises due to a translocation, the Philadelphia chromosome, that occurs in a blood stem cell. It is the first cancer that can be treated with molecularly-targeted chemotherapy. This therapy, imatinib, blocks the abnormal effects of the Philadelphia chromosome and thus only inhibits cancer cells. Its discovery was a great success in cancer research because it allows for the first time a specific killing of cancer cells instead of a poisoning of all dividing cells in the body, as conventional chemotherapy does. We did an analysis of CML kinetics in vivo and could establish that imatinib does not inhibit cancer stem cells and hence will not lead to a cure.
Figure 2: Franziska Michor's book (2006)
Q: A primitive question about the mutations. When the DNA mutates, how much does it differ between the different cells? Is it a good approximation to talk about a single DNA code for the whole organism throughout its life? And another question: is there some qualitative difference between the "bad" mutations (such as those causing cancer) and the "good" mutations that were necessary for the humans to evolve, for example?
A: In humans, the mutation rate per base is about 10-10 per cell division, and the human genome consists of 3 billion bases, so it's less than one mutation per cell division. Furthermore, most mutations will not have an effect on the protein (they are silent) because of the redundancy of the genetic code, the presence of non-coding DNA etc. So it is a good approximation that the genome remains more or less constant throughout life.
Concerning the qualitative differences between "good" mutations - those that lead to higher life - and "bad" ones - those that lead to diseases and cancer: one of the perils that had to be overcome in the evolution of multicellular life was to make sure that individual cells would cooperate in the endeavor of multicellularity and not revert to uni-cellular, selfish behavior. The evolution of growth factor control, cell-cell signaling, tumor suppressor genes etc. contributes to this goal. Cancer can be regarded as reversion of such cooperation within a tissue: a cell receives a number of mutations that deafen it to intra- and intercellular control to the end of dividing for its own benefit and to the harm of the community (tissue). So only if the many parachutes build into our cells to rescue cooperation are destroyed, cancer can evolve.
Q: Do I understand well that your current work is theoretical in nature? Did you ever have to kill mice? What do you think about the hardcore experimental biologists and what do they think about you? Which group is more responsible for the actual victories in the battle for the cure for cancer? A question that is also relevant in high energy physics is: how many important things do you think you can learn by general mathematical methods without doing the detailed experiments?
A: Right now my work is purely theoretical, but I did do experiments during my undergrad in Vienna. I never had to kill mice, though. I think biology without mathematics is like physics before Newton: you could observe a star's position yesterday and today, but you could not predict where it would be tomorrow. Biology without mathematics is phenomenological - we are merely collecting data, but we cannot make useful predictions or obtain a quantitative understanding. Biomathematics is still in its infancy, but has already and will make important contributions to biology (Hodgkin-Huxley etc.). However, a model is only as good as its assumptions. We do of course rely on experimental input, and models and their predictions have to be testable.
One might wonder why it took so long to introduce mathematics into biology. I think it has to do with the high complexity of biological systems - while you can do isolated physical experiments with, say, lasers without worrying about rain, biology does not work like this. We cannot study one signaling pathway in vivo without having 10,000 other chemical reactions occur simultaneously.
Q: It's a well-known fact that cancer involves a lot of mutations. Are they just a by-product, or the primary reason why cancer occurs in the first place? If they are the reason, what starts the mutations?
A: Cancers are caused by mutations which are caused either intrinsically - the replicative machinery copying DNA makes occasional mistakes - or extrinsically, such as by UV light or carcinogens. If the first random mutation hits a gene that normally prevents genetic instabilities, then an increased mutation rate can drive the whole process, as discussed above.
Q: According to your knowledge and beliefs, what are the main effects that may cause cancer and what is the best recipe to avoid it? Would you say that most of the information in the media about new carcinogenic effects is trustworthy? How easy and reliable is it to check whether a particular factor accelerates or inhibits cancer?
A: There are some well-known carcinogens such as radiation, UV light, asbestos, benzpyrene (in burnt meat) etc. These should obviously be avoided. New carcinogens can only be validated by their immediate effects on cells and by statistical analyses of cancer incidences in people exposed to such agents, as done in cancer epidemiology.
Q: Finally, the questions about the innate aptitude of women have influenced the recent life at Harvard University a lot. Would you agree with The Reference Frame that women have simply been optimized for different things than men by millions of years of the human evolution - for example, they are not as good as men in driving trucks?
A: Fortunately, you picked a good example. When I was 18 and was getting my driver license, we decided that I also needed a driving license for the truck. It may be fair to say that my driving may have been better than what the boys showed, with a possible exception of the acts that require a lot of physical force, such as connecting the truck to the trailer - but these shortcomings can be removed by more careful driving or targeted kicks.
In general, I do agree that evolution has led to different optimizations and specializations in men and women. Men might be better in hunting, fighting and wood-cutting while women may well be more talented in cooking, child-raising and berry-collecting. However, scientific research is such a recent occupation for humankind that evolution cannot yet have had any effect. It is possible that the two sexes have different abilities in jobs which derive from stone-age occupations; women might be more able chemists (if you think that chemistry is derived from cooking) whereas men could be better architects or army generals. But I strongly question the evolutionary pressure on the ability to do mathematical biology or drive trucks.